<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Rorschach on Medium]]></title>
        <description><![CDATA[Stories by Rorschach on Medium]]></description>
        <link>https://medium.com/@0x_Rorschach?source=rss-0b668e3fef6c------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*lHZqKPSuGVAl-XeKtgQm0A.jpeg</url>
            <title>Stories by Rorschach on Medium</title>
            <link>https://medium.com/@0x_Rorschach?source=rss-0b668e3fef6c------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 16:52:16 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@0x_Rorschach/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[Exploring the Scientific Method: An In-Depth Examination of Methodology, Algorithms, and Empiricism]]></title>
            <link>https://medium.com/@0x_Rorschach/exploring-the-scientific-method-an-in-depth-examination-of-methodology-algorithms-and-empiricism-947109b4913a?source=rss-0b668e3fef6c------2</link>
            <guid isPermaLink="false">https://medium.com/p/947109b4913a</guid>
            <category><![CDATA[empiricism]]></category>
            <category><![CDATA[science]]></category>
            <category><![CDATA[algorithms]]></category>
            <category><![CDATA[methodology]]></category>
            <dc:creator><![CDATA[Rorschach]]></dc:creator>
            <pubDate>Sat, 14 Jun 2025 11:43:48 GMT</pubDate>
            <atom:updated>2025-06-14T11:43:48.428Z</atom:updated>
            <content:encoded><![CDATA[<h3>1. Introduction to the Scientific Method</h3><p>The scientific method is a systematic way to gather knowledge and test an explanation for specific things. This inquiry-driven method has been widely applied in various fields, including physics, chemistry, biology, and social sciences, and is regarded as an essential component of engineering design, evaluation, and analysis. In this method, an initial question leads to the formation of a plausible explanation, known as a hypothesis. This hypothesis is tested using experiments, analyzing data, and making conclusions that lead to the acceptance or rejection of the initial hypothesis. Although rarely followed as rigorously as described, the scientific method is a guiding principle for most scientific inquiries 1. This thorough examination of the scientific method discusses its compliance-generating process, the dependence of methodology on discipline and problem complexity, current advancements in scientific inquiry, and opportunities for better matching of analytical tools to experimental approaches.</p><p>The invention of the scientific method can be conceptualized in two steps. The first step is compliance to the formalized process of inquiry. Compliant thinking is a process by which a well-planned step yields an answer, leading to refinements, steps, more answers, and further refinements. The second invention takes place within the context of the compliance-generating process and is highly contingent upon the former (inducing gradual changes in the process and subtle departures from the norm). This gradual transition took place across the Northern European continent during the rise of universities and was accompanied by the invention of novel thinking apparatuses and novel teaching practices. Methodology depends on both discipline and the complexity of the problem addressed. While physicists commonly use mathematical formulation as methodology, mathematicians primarily disarm problem complexity. Other disciplines kill complexity by restricting experimental considerations or heuristic exploration. Given the complexity of the procedures, it is not a small task to forge methodologies where discipline complexity suddenly heightens up. Compliant inquiry is painstakingly slow, but it generates reliable results, rendering comprehensive differences between disciplines with regard to model reliability. Modelling-methodologies may be developed for compliant inquiries, but this modeling takes far more time and effort, which has proverbially been acknowledged by a design engineer.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/468/1*rZIbT8c5ei9Xn-QG60Fg0g.png" /></figure><h3>2. Historical Context</h3><p>The scientific method, first formalized in Europe in the 17th century, has shaped the way the natural sciences develop knowledge — it has become implicit that scientific results must be achieved using this method. The recognition that there are tacit principles of scientific knowledge generation dates back much further, and in fact all cultures have alternative methods of knowing that can be labelled science. As a consequence of these differences in basic epistemology, there are formidable problems in cross-cultural science, notwithstanding the vastly different histories and institutions of the natural and social sciences. This incertitude and dissatisfaction has led some writers enjoying the prerogatives of having “modern” knowledge to urge developing countries to “catch up” with the methods of the natural sciences 1.</p><p>Modern science as practiced in Western Europe imposes a number of constraints on the generation of knowledge. These constraints fall into eight broad domains that define the natural sciences — the units of knowledge, types of empirical observation, a hierarchy of empirical test, the relationships between knowledge units, underlying metaphysical assumptions, values, social structures, and the production of knowledge. Only through scrutinizing each of these in turn does the inappropriateness of basing natural knowledge on method become evident. It follows that because there are many different methods of generating knowledge, many different sciences are valid. However, this pluralism does not remove the problems of weak links, and so a science regulatory system must be developed, one that is culturally neutral and allowing for competition between systems to provide the best science.</p><p>The scientific method is predominantly fluidic and testable. For a proposition to be scientific, it must be explicitly stated in terms of a mathematical model, which serves various roles, from being a link to the physical world to providing many possible approaches for testing and revision. Generally, with a good model, the necessary equations can be set up by engineering cleverness, and results can be explored by applying rigorous analysis or by using computational tools. Importantly, these strategies impose very strong constraints on the types of question that can be posed, thus controlling chaos. Hence, methodologies of this kind can be bundled into the category of “algorithmic thinking.”</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/792/1*WtN3JbTSCurfr5tRaW3RTw.png" /></figure><h3>3. Key Components of the Scientific Method</h3><p>The scientific method is biased toward the biblical way of thinking “in small steps.” This is one reason for the “one step at a time” mindset that was impressed in the very young. The famous quote illuminates the problems that result from taking well-planned steps forward all the time. This is the point at which the logical theory usually gets stuck. With the advance of high-throughput data generation, combinatorial experimentation, and big-data analysis, scientific effort has spontaneously shifted into exploration. A brief account is provided on the latter approach as it applies to understanding biological systems. Nevertheless, it was frustrating when not knowing whether an experiment would be insightful, at least within some frames of reference. Subsequently, the guiding principles of the discovery-time investment management in the explorative scientific method were discussed 1.</p><p>Nature has its own rich and diverse internal problem-solving schemes or strategies for assimilating information, and these are conceptually at par with human-designed algorithms, some of which were discriminatively or randomly invented by scientists. The choice of the right screening or inference method in biological research is generally influenced by fashion or discipline. Ten such strategies are discussed here that are already widely used or will likely soon be implemented. Various classifications of types of scientific strategies are possible, and an arbitrary number may be useful and provide some information to persons from different backgrounds. A degree of redundancy is therefore unavoidable. Each method is supplied with a brief historical outline, indications of applications, a discussion of the general rationale, perceived advantages and caveats, and especially lessons from their use during the design of the explorative biological studies before the start of the sequencing project in the late 1980s. The final short section sides with a somewhat perplexing discovery made in casino gaming by a mathematician who eventually won a few million dollars by spotting the house advantage in gambling games.</p><h3>3.1. Observation</h3><p>Recalling our definition of observation, the set of observations which constitute a cover is inconsistent, and the union over the set of possible configurations of the universe of discourse associated with these observations includes all admissible configurations or models. Denote by R the set of all relevant models. For any cover E of the set of observations O, denote by K the set of models satisfying O, and K the set of models satisfying E. It can be shown from the properties established above that A is one of the configurations R unless A is a collection of new observations in a inconsistency. The term ‘observed’ must therefore be understood as meaning that at least one of a given set of relevant observations has been made 2. The outcomes of the experiment correspond to observations and are not directly related to the observer’s theories as expressed by his language. For any given possible outcome ai and associated cover Ei and collection Da of additional observations or propositional functions, it holds that P ( ai | Do ) = P ( a | Do EtEi ) = P ( ai | Do EtEi ). The event E, which consists of the outcomes of the experiment, is no longer confined to a single a priori probability assignment. Given a prior probabilities P, the question which may now be asked is of the form P ( o | D O EtE ). Denote the distribution p ( o ) = P ( o | D O EtE ) of the outcomes given the observations outside E (with the prior assignment P ) by P ( o | D O EtO ). After the experiment has been performed, and one of its outcomes ai has been realized, the observations associated with that particular outcome are no longer general. Rather, the outcome ai is one of the events ‘o observed’ Sti and it is to be expected that, whatever a priori probabilities P , these events can be assigned a posteriori probabilities P’ in such a way that for i # j, P’ ( ai ∪ aj ) = 0. In other words, one has a conditionalization of posterior probabilities.</p><h3>3.2. Hypothesis Formation</h3><p>As science advances, mass knowledge can be veraciously acquired. However, the endeavor to transforming observations and mass knowledge into hypothesis is as demanding as the prior endeavor. The next two sections address this challenge. To show that scientists can see a certain aspect of circumstances unnoticed by many others, which is commonly disregarded as trivial or uninteresting will be one aim. To examine the process of hypotheses formation used by others, which is regarded as the most essential part of science, will be the other aim.</p><p>These two aims imply different aspects of similar questions: how can propositions be anticipated without observations? To show how these two aspects are related will be the first task 3. To accomplish this task, a simple but significant experiment on observations and hypotheses formation will be analyzed. This will illustrate that it is possible for one person to see things that are not obvious to others. This also implies that circumstances have an observable aspect while hypotheses formation has a causal one. By repeatedly looking at the same circumstances from different positions or by viewing different aspects of the same circumstances, concepts that can never be exactly observed can be found. It seems that there exist some states on which there are no properties. A certain aspect of properties of states is considered as hypotheses and an endeavor to find accurate properties of the corresponding observations is regarded as observations 4. Thus, an effort on hypotheses formation is not an isolated attempt.</p><p>However, it is difficult to find justifiable reasons for this effort. It is too easy to take certain properties of one observable as the hypothesis of another observable that must have properties but can never be exactly observed. To avoid this, the following three aspects of hypotheses formation will be discussed. The first is how to formulate a hypothesis satisfying certain properties as widely accepted hypotheses formation methods. The second is whether a hypothesis should contain some ignorance as a trade-off of generality and possibility as capacities to do something. The final one is whether one can find an accurate observation to verify a guess about a hypothesis.</p><h3>3.3. Experimentation</h3><p>Beyond the familiar setting of testing established hypotheses, experiments are deployed for a variety of purposes. They can be used loosely to explore the consequences of certain theoretical assumptions; they can supply theoretical posits from which broader generalizations can be drawn. Experiments can be open-ended and exploratory, often referred to as ‘situated curiosity engineering’ 5; they can be conducted to develop representations of phenomena and establish idiographic generalizations — to tell stories about singular objects or events; they can serve as models of larger systems; and they can also be used to provide evidence for cover stories in the absence of a more suitable theoretical framework. Attempting to catalogue the range of things that count as experiments, but few would doubt that exceptional cases exist. At the other extreme, overly radical definitions of experiments run the risk of including a wide swath of activities such that nothing is said of interest.</p><p>Interestingly, exploration does have a precise, mathematical definition in terms of information-theoretic measures of surprise. In terms of such criteria, modeling could be construed as exploratory in some cases, and even so-called purely observational cases would be exploratory in the same sense, suggesting that a single unified understanding of exploratory experiments is unlikely to be forthcoming or even desirable. Nevertheless, there are a number of examples that sit awkwardly in the margin of standard understandings of experimentation and that stand in stark contrast to models or purely observational cases. This exciting range of empirical practices points to ways in which philosophy can benefit from crossing disciplinary boundaries to examine how scientists actually structure their investigations. Such work must also aim to be interdisciplinary, studying the cultures and practices that scientists build up and so that more mundane practices, and the rationales behind these, are also examined.</p><p>In closing, pointed to a number of fruitful avenues of exploration. The first of these was to take a close look at cases of situated curiosity engineering as they play out in and around the laboratory. There is a rich tradition in science studies of uncovering the details of everyday practices: these successes can be described and, ultimately, new insights can emerge where the holes in existing pictures of experimentation are found. The second avenue was to explore the range of idiographic generalizations that experiments in the empirical sciences do yield. Many have been at pains to argue that most often these generalizations are ‘difference-making’ in form and not universal — devoting careful attention to examples.</p><h3>3.4. Data Analysis</h3><p>Data analysis and information extraction is the second empirical methodology phase where a scientist produces models that generate theoretical output, and empirical data is analyzed concerning the model’s predictions. The goal is to mathematical models to explain the empirical data through the process of graphically or algorithmically recognizing or reproducing regularities in the data. Statistical analysis is what scientists rely on to perform data analysis. The underlying motivation is to statistically distinguish the systematic findings from those due to chance. These methods assume modeling the underlying distributions of the datasets, e.g., Gaussian or Poisson distributions for continuous and count data, respectively. Accordingly, the analysis is performed, and the relevant findings are interpreted under fully parametric assumptions. How valid or meaningful the analysis of the data is, depends on how realistically the assumptions hold. Being unable to apply any analysis technique, or more generally, any model inconsistency with the assumptions means losing all the information extracted so far. On the one hand, most data-driven processes implicitly assume distributional wise handling of the data via (multi-variate) generalizations or embeddings. On the other hand, nevertheless, the comparisons between alternative methods in hand are practically blind, since all the analyses are performed with few rather fragile routines. Finally, addressing and quantifying knowledge extraction from the data means freedom from distributional fallacies, as well as quantifiable universality to extract knowledge from raw data. In this aspect, a completely novel framework is proposed for performing data analysis without distributional assumptions. To this aim, a simple mathematical formalism will be motivated through which to consider valid inferences to the data. The very notion of information concerning some system will be inquired, leading eventually to a consistent approach to represent all kinds of information. A formal understanding of what kinds of knowledge are distributed concerning the data will begin from the data, namely how much understanding mentionings of the data emerges by the data 6. It will be shown how the latter is informing better structures concerning the data, which w.r.t. to the way of performing analytic comparisons 7. Finally, the concept of locally-informed models will be introduced, to analyze the data with respect to the knowledge generally produced by other lenses.</p><h3>3.5. Conclusion</h3><p>The understanding of algorithms, methodology, and empiricism is a powerful toolkit for scientists and critical thinkers alike, but one which is often overlooked in science education. It is hoped that this text will help further the discussion behind this understanding as it is applicable to the vast variety of works and ideas that can be termed as “scientific,” and help bring forth an important paradigm shift in the common understanding of “science” to be more inclusive, critical, and thoughtful. The scientific method has been used by the human race to advance broadly almost every aspect of human existence. However, there is a dearth of literature, including popular literature, in understanding and explaining the nature and advantages of “scientific” methodology 1. The scientific method and scientific philosophies are broad and deep, and it is hoped that the currently presented exploration and discussions divulge some tidbits which may help inspire the next generation of thinkers. The hope is that engineers, artists, philosophers, and psychedelic travelers alike will all find something in here to help further their work as it pertains to human imagination and invention.</p><p>As new methods are developed, understanding and educating others about them becomes an important thread of the scientific method and paradigm of modern society. There is much similarity between the development of new scientific methods and the development of newer mathematical algorithms. In both cases, meticulous stepwise logical deduction results in new truths, whether they be new mathematical equations or heuristics, or new data generation techniques, data-fitting methods, or environmental detection, and engineering techniques. However, it is important to note that within this deduction process any of the carefully observed and related data pertaining to the model inputs or initial values need not be justified or elucidated 8.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/411/1*NL0yjlCvQCytgW2oxd7dfg.png" /></figure><h3>4. Methodological Approaches</h3><p>The traditional basis for the scientific method rests squarely with actions that lead to the formulation and testing of a scientific hypothesis: that is, scientific inquiry. A scientific hypothesis is usually conceived in a heuristic fashion based on a broad range of experiences and knowledge. Such intuitively derived hypotheses require the thoughtful consideration of a number of varied influences or factors, both internal and external, whose effects on a measured parameter are postulated. Constructing a scientific or mathematical model, then, is the act of formally representing both pruned transmission pathways from hypothesized influencers to a measured parameter and assumptions regarding the spatial and temporal ranges of these included influences. In contrast to the crafting of scientific hypotheses, mathematical or computational models can be elaborated as a deductive exercise to mirror a measurable attribute with no precise preconceptions as to the components or structures that might need to be included. It is only in later iterations that scientific and mathematical modeling might engender duplicate or concordant expressions for a biological problem. Certainly, persistence in iteratively creating mathematical models without scientific justification is fruitless, a mathematical model without scientific rationale being more of a toy than anything else, but these two distinct methodologies could be employed separately yet yield very similar results. Given this perspective, it is not altogether aspectious to begin a verbal exercise with, “Once upon a time, in a large biological system far away…,” because that supposition contains no scientific or mathematical rationale other than “Because it could be.” The impossibly rich biological interactions expressed by such a murky heuristic sketch are sufficient to make any reasonable model a reasonable guess. The gregarious pursuit of mathematical elegance is not science, but an erudite exploitation of a powerful and flexible toolset. By disassociating model formulation and model use, the strategic theories expressed transfer both the excitement and the trepidation exhibited in such unconditional modeling insights into the realm of scientific inquiry. Despite presenting its own terrific challenge, acknowledging that strategical selection of the scientific method is of equal importance to and utterly different from technical issues opens a wide avenue of further insight into the intricacies and intricateness of biological systems.</p><h3>4.1. Qualitative Methods</h3><p>Quantitative research often induces automatic mental action, particularly when working with numbers, with the frequent publication of quasi-experimental studies automatically inducing action by critical appraisal researchers. However, evaluating elaborate qualitative studies with naturalistic settings, co-construction of knowledge, and comprehensive analyses is a labor-intensive, time-consuming endeavor that requires careful reading and re-reading 9.</p><p>Researchers should be conscious about critiques of qualitative methodology. The key point is that qualitative methods are not subjective, amateur, artefact-focussed, or less useful than quantitative research methods. These ideas about qualitative research only have power because of unquestioned assumptions behind them and the worryingly big analytic blind spots they leave in their wake. Much qualitative research has continued to go unquestioned, and as a result the vague idea that qualitative methods ‘are subjective’ can sometimes inappropriately enter into debates in scientific disciplines. As a result, less-than-rigorous qualitative research designs and analytic procedures do sometimes exist and are sometimes published 10. Nevertheless, what counts as rigorous, and the procedures used to achieve that rigor, has to be thought out in relation to the conventions and norms of the specific scientific discipline.</p><p>To explore and illustrate questions regarding the nature of qualitative methodology. Discussions of methodology often fall back on an epistemological analysis of the quantitative-qualitative dichotomy. This is unfortunate because it serves academic debates rather than the practice of qualitative methodology. Instead, the discussion areas focus on the implications of methodology for practice. To make different qualitative methods comparable and tell the academic community which qualitative methods are the ones to use when aiming to address which questions. To contribute to the literature documenting distinct methodologies for qualitative research. To focus on the question of data transformation.</p><h3>4.2. Quantitative Methods</h3><p>Quantitative method systematically investigates phenomena by collecting quantifiable properties for analysis. A case study utilizes this method to explore online collaborative learning. Qualitative methods examine the issues at a deeper level, providing information regarding feelings and opinions. The mixed-method methodology combines both quantitative and qualitative data collection. A case study examines which data collection method produces better data quality in terms of nonresponse.</p><p>Quantitative and qualitative are fundamental approaches to scientific research. Mixed-methods combine approaches to expand and correlate knowledge. Here, approaches differ qualitatively through diverse data types, analysis methods, and paradigms. Quantitative, qualitative, and mixed-methods are the target methods. Quantitative research begins the research design process with a model involving quantitative variables. Variable relations generate research questions regarding overall prediction, comparison, or difference. Informed by theory and prior study, researchers hypothesize specific variable relations. Hypotheses postulate difference and correlation, express directionality with two-tailed non-directional prediction, and undergo tests by measurement with instruments. Testing validity encompasses design, content, reliability, construct, and criterion. Research findings include empirical observations or data matrix hypotheses testing combinations. Results link findings to hypotheses, providing direction for discussion and conclusion comprehension. Review interpretation should revisit findings, questions, hypothesis, and assessment procedure or design weaknesses 9.</p><p>Result formats range from scientific observation and inferential statistics summary to exploring and diagnosing implications analysis. Analysis begins analysis on a variable basis through unobserved data modelling. Planning issues double models using grouped states of a data matrix for paradigms or stages. State variable analysis accounts for fixed effects with mixed unobserved data. Unobserved grouped states generate independent sample discovery, selection, exclusion, and dependent states. Fixed and random quantifies include detected states such as random effects and error variances linking with sampled dataset frequency distributions, regression coefficients’ distribution, and unobserved variable category probabilities.</p><h3>4.3. Mixed Methods</h3><p>Research questions are not always answered using quantitative or qualitative approaches only. Some research questions can only be answered using a combination of both methodologies. Mixed methods combine the strengths of qualitative and quantitative methodologies. In the social sciences, qualitative research has been used to explore a phenomenon, while quantitative research has been used to test hypotheses and theories. However, in the past, there were concerns about the factors assessing trustworthiness with qualitative algorithms: the acceptance of numeric or mathematical treatment created resistance on the part of qualitative researchers conflicting with a numeric system. Although the acceptance of qualitative and quantitative research as different paradigms of investigation or ontologies and epistemologies is waning, even today, few researchers use combination methodologies. Because quantitative methodologies have been formalized and widely accepted, many Dutch Institutes in the social sciences support research using quantitative methodologies.</p><p>This study seeks to clarify the different approaches and alternatives available for those wanting to use mixed methods in their studies. It intends to increase knowledge about formulating and executing mixed method research. A literature review on mixed methods is performed, recognizing predominant authors and identifying some mixed methods approaches. Then the methodology of the study is presented. Finally, the results are analyzed and discussed to analyze the impact of each mixed methods design, characterize the steps included in each approach, and explore the benefits and limitations associated with each approach 11.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/852/1*GDXkcpnzAQGsTn-LwKrRpA.png" /></figure><h3>5. Role of Algorithms in Scientific Research</h3><p>An algorithm is a specific class of technological rules. A technological rule is a well-defined and implementable rule that prescribes how to transform not necessarily computable but, at least, a finite state data into some output or data. An algorithm is a well-defined and implementable sequence of computational steps composed by some operational rules that transforms some (often infinite) input into some output (finite or infinite). Computable rules are a subset of algorithmic rules. In general, algorithms build on some machines or computational systems such as Turing machines, automata, registers, etc. The existence of these abstract machines is the basis for results on the computability or tractability of the corresponding procedures. Such results are profound and transcendental. However, they are often found rather abstract, formal, and system dependent. Thus, when they are used in applications of mathematical or computer sciences they have to be transformed and modeled into more concrete building blocks such as programming languages or data structures. This is the realm of computer science and algorithm design in particular.</p><p>Research on algorithms is as old as science and mathematics. Nevertheless, it is considered to be a specific domain of research in computer science where both mathematical and programming knowledge and skills are involved. In the domain of computer science it is related to a whole set of disciplines (such as algorithm analysis, numerical analysis, algorithm design, programming, etc.). Within computer science it has its own methodologies for algorithmic analysis, design, and computation. On the other hand algorithms have also drawn the interest of research in linguistics and logic. More completely, algorithm design is a specific class of technology and capital goods produced by specific activity systems, namely the algorithmic activity system. There are different definitions and descriptions of algorithm design. Algorithm design has to be performed in consideration of the available tutors, materials, and the intended consequence. It involves deciding the learning target and elaborating a pedagogical methodology. An algorithm can be or become a computer algorithm.</p><h3>5.1. Definition of Algorithms</h3><p>A wide variety of approaches have incorporated mathematical models and methods in sports analytics, and there are various different ways to classify those approaches. One way of categorizing them is to use a taxonomy in terms of the mathematical methodologies or redresses used. Therefore, a fundamental, general and very broad definition of methodologies is essential 12. The term ‘methodology’, being a compound word of the Greek roots ‘methodos’ (stream, path, way) and ‘logos’ (reason, science), first appeared in the early 19th century 13. Basically, methodology denotes the perspectives, modes, tools, and processes used in pursuit of knowledge. In the social sciences, methodologies also refer to code of ethics in how to conduct research. In this present context of sport analytics, it refers to the (mathematical or computational) modelling or algorithmic methodologies, which in turn can also be categorized based on their paradigms, ontologies, processes, platforms, and products.</p><p>A model is a conceptual representation of an entity, a phenomenon, a system or a problem. A model is mathematical in nature if it is described and manipulated using mathematical expressions and relations. A model has a number of specific features, including its purpose, assumptions, simplifications, representations, and levels of detail. A solution methodology is an algorithm or a computational process that operates on the mathematical model and transforms them into a solution. An algorithm is mathematically defined as sequence of well-defined, unambiguous, and recursive operations or steps to obtain an output from an input or to solve a problem. It may involve initialization, termination, and decision-making.</p><h3>5.2. Types of Algorithms Used in Science</h3><p>While theorizing is clearly part of science (to come up with hypotheses to test), there is also a component of generating numbers, with formal methodology for testing given the numbers. The core part of this methodology is based on algorithms, putting science right along with computing in which algorithms (and their complexity) are a central issue. Variations have long been used to describe an algorithm class: method, method family, algorithm family, algorithmic family, algorithm procedure, machine model, procedure family, and programming style. The emphasis here is on classes of algorithms that share some characteristics, not on particular algorithms, so algorithm class is more appropriate.</p><p>There is an ongoing debate in computer science how algorithms should best be studied. Although there are arguments in both camps, both views can be integrated into a broader framework, which is general enough to encompass a variety of approaches. Algorithm engineering is the process of transforming a class of algorithms into an implemented software system, and ensuring that the final product is of high quality. Algorithms are more than a mere mathematical model; they are the central element of modern computing and information technology. Because algorithms have become ubiquitous, ensuring the quality of the software implementing them has become very important.</p><p>As computing devices become faster and more intricate, the quest for faster and more intricate algorithms has become possible and desirable. A research area was created, devoted to mathematical theories and new algorithmic techniques for improve algorithmic performance. Research on algorithms has resulted in important theoretical results in various fields, and increased understanding of the interplay between efficiency, performance and quality 13. In fields such as computer graphics, computer vision, computational biology and computational geometry, the efficient algorithmic solution of complex modeling, interpretation and design problems has proven so successful, that these areas have become mature scientific disciplines of their own.</p><h3>5.3. Algorithmic Bias and Ethics</h3><p>From movie recommendations to friend suggestions, algorithms are ubiquitous in daily life, showing results based on the underlying patterns found for given inputs. An algorithm is a set of systematic instructions that can solve problems and produce new information; it can be articulated algorithmically as a function, resulting in a mapping from input to output 14. Algorithms can be deterministic (when given a certain input, a unique output is seen) or probabilistic (when input data is submitted, the output is produced with some probability). Algorithms can take many forms, for example, they can draw pictures based on user input, seek to play a zero-sum game, or identify the tone of a text. The combination of the ability to handle a very large amount of high-dimensional input data and measurement precision has made machine learning the dominant approach to algorithm design in practice, yet this might render them impossible to completely understand.</p><p>As a synonym for a recipe, the word is self-explanatory: given the input products and parameters, along the recipe instructions, the output food item is obtained. Algorithms play a very central role in achieving many modern quotidian efficiencies. However, they are often unacceptable for many applications, especially when new and/or incorrect outcomes surprise people. Due to their role in major events, some infamous algorithms have attracted public scrutiny in the high-profile media. As they have been shown to be affected by systemic biases and prejudices, algorithmic fairness has become an active research topic in many disciplines, especially within computer science, social science, and law.</p><p>Though some early works expressed concerns about algorithms potentially legitimizing innate discrimination, biased effects due to locale and usage context attracted extreme attention in mid-2010s. Algorithms exert a wide range of impacts on daily life; they are important in content moderation for social media platforms and in deciding whether a defendant is eligible for bail on judicial platforms. The ubiquitous presence of algorithms makes it more pressing to understand how people think of them. Though human conceptualization of various concepts, such as likelihood or type, has long been studied, very little research has been conducted to understand how non-experts think of algorithms.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/792/1*v9dK67ADXHGP8w7AavCavA.png" /></figure><h3>6. Empiricism in Science</h3><p>The scientific approach typically consists of observation, measurement, and experimentation 15. Observation helps the researcher to formulate essential questions about a phenomenon under study, to build models, and to derive hypotheses that can be tested through experimentation. Measurement is essential for both observation and experimentation. A scientific hypothesis must be refutable to be meaningfully tested. Tested hypotheses are compiled and communicated in the form of laws or theories. At the heart of the scientific approach are research methods in general and the empirical method in particular. Empirical methods leverage evidence obtained through observation, measurement, or experimentation to address a scientific problem. Evidence should be based on qualitative and quantitative research. In particular, they should be obtained through careful planning and recordings of the methods and results. To perform scientific research in software engineering, one has to understand the available research methods. Since the methods are extensive, a structured overview of the family of research methods and their links to typical research questions in software engineering is provided. The scientific method is inductive and tries to extract from the world some model that can explain a phenomenon and to evaluate whether the model is representative for the phenomenon under observation. When exploring software engineering (SE) phenomena, the researcher has to be careful about sampling. The empirical method normally follows an iterative and incremental approach that can begin with an exploratory survey, followed by case studies in an industrial context to better understand specific phenomena and controlled experiments to investigate cause-effect relationships. In this procedure, theories are iteratively constructed and refined. Despite being useful for research in SE, formal models are not explanatory but only descriptive, and therefore in turn not a viable approach for scientific research. The scientific method, inspired by natural science, is traditionally used in applied areas. Scientists mandate to understand a phenomenon and to produce general knowledge about it. The applied knowledge is then transferred to the engineering community to construct new products, processes, or methods. The engineering method is dominating in industry. It is deductive and constructs a model that satisfies given constraints in the best way, navigation technology being an example of this search-for-optimization approach. The embedded scientists also do research and therefore both methods are used in parallel in a co-dependent way.</p><h3>6.1. Definition of Empiricism</h3><p>Empiricism is a theory of knowledge that asserts that knowledge is derived from sense-experience. Empiricism emphasizes the role of experience and evidence from the senses in the formation of ideas rather than innate ideas or traditions. It relies on data collected through experimentation, observation, and documentation of events. The term empiricism comes from the Greek word empeiria, which means experience. The theory is sometimes understood broadly to include all knowledge as a product of experience. More commonly, however, empiricism is understood more narrowly as a theory of justification.</p><p>Empiricism is often contrasted with rationalism. Empiricists typically reject rationalism’s claim of synthetic a priori reasoning. A widely accepted standard for defining empiricism includes summarized definitions: (1) “All knowledge is originally derived from sense experience.” (2) “The primary laws of all science are generalizations of a vast number of similar facts observed or produced.” (3) “Except in the case of mathematical truths, no inference of any kind is theoretically certain.” (4) “All probable truths are approximations to the truth.” (5) “The degree of probable certainty depends mainly on (i) the extent of the induction, and (ii) the complexity of the facts.” Empiricism is founded upon the assumptions of a physical reality that exists independently of each person’s knowledge of it, the reliability of the senses to tell observers what that reality is, and the belief that past experience is relevant to future experience.</p><p>The root assumption of empiricism is a belief in an objective reality. The reality exists independently of subjective experience. This requires the existence of physical assumptions and laws, an external world, a subjective mind, and a reality that is not created or influenced by a perceiver’s knowledge. There is an assumption that there is a reality that remains constant no matter who is looking at it. That observers draw the same conclusions about reality in experimental settings suggests there are truths that are independent of all people’s biases, opinions, beliefs, or experiences. The scientific method is based on this belief of an objective reality.</p><h3>6.2. Empirical Evidence vs. Anecdotal Evidence</h3><p>George Gaddini, ALCHEMY, a combination of encyclopaedia and manual-like collection of methodical matters and elaboration methods dealing with education, learning, work, and entry processes offering an unabridged declaration of all semantic constructs taken into account. If manipulation is absent, the user is left with semantic means borrowed from someone else unshaped up to one’s own need, a burden stemming from a blasted intellectualisation. Misuse then simplifies to a trivial ill-behaving by the user adapted to the purpose of defence, where an excuse for lying, deceit, woeful training, and foreboding are so readily available. Formalism wherein explication of the first ought to ground out decent offers relevant for educational training matters so far as methodical thought is concerned to go beyond first. Appropriateness of thought and action both cannot be of a kind within a formalism that is boundless, since the semantic space it defines is vacuous. The quest on behalf of educational psychopathology is to describe the section in which many trains of thought and action ‘cease to be pertinent’, and the things that such inappropriate conduct have to do with one another.</p><p>Corruption of principles comprises violation of Gaddini’s notion of nestled heuristics: a typical example is the emergence of a mainstream spelling book producing enormous error as for formal and content-pertaining characteristics; another by teachers elaborating Gaddini’s altres tempus assumption wrong the that systematic development to get the grip on unconscious projection could be bridged by high-pitched donation of drawing on the black-board a bit every day. The voracity with which contemporary experts employ Gaddini’s neglect of chaos, contamination, and the continuous conflict unarguably engender educational maladies beyond those whose nature were formerly imagined. The modal credance which ought to have enabled scientists to avoid prescriptive and turn early and preventively to pre-emptive, preventive, and speculative explanations right away is that over-conference. The notion of consequence-cause boxes is employed to shed light upon the talked about pirouettes how such scientific work nowadays is commonly conducted: deliberation space is an immensely high-dimensional one wherein lots of tuneable knobs have to be set just-right with formalism located at a much higher level. The plight of epistemology is concerned with ‘how any knowing of various kinds is possible, and what features of constructions must on questions encircle the conceivable scene’ which Gaddini’s gold-fingered principles will steep as an unseen beautification for posterity.</p><h3>6.3. The Role of Empirical Testing</h3><p>The main criterion for the ‘truth’ of scientific models is their empirical testing. A model’s validity is typically associated with its ability to ‘pass’ an empirical test. The notion of an empirical test is related to the notion of a ‘testing method’ or, more generally, a ‘method of verification’ . According to some accounts, the goal of an empirical test is to find some discrepancy between the model and some observable data; a discrepancy typically implies that the model cannot be right. Other accounts tend to focus on measuring some quantity. Eventually, results are usually interpreted with some prescribed rules — such as hypothesis testing or ‘confidence intervals’ — in order to determine whether the data have passed the test or not.</p><p>Nevertheless, the literal notion of an empirical test is impossibly broad. Every mathematical experiment is an empirical experiment; every numerical simulation of a model is an empirical experiment. The testing of hypotheses in terms of empirical data is nevertheless often thought of as scientific testing, as distinct from mathematic processes 16. Yet even this narrower notion seems still overly broad. For example, the many numerical simulations of the trends of the temperatures of rooms are mathematical models but one would seldom think of them as scientific. In the case of scientific models, the historical development of scientific understanding suggests to some that a proper empirical test is of the model involving a ‘testing methodology’, a data-generating process which is theoretically ‘independent’ of the model under consideration. If the model generates a ‘predictive expression’ for a quantity which can be observed or measured using a specified empirical procedure, then this quantity is said to be empirically tested.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/912/1*cDZwIdyUQ8eZocf76EMQuQ.png" /></figure><h3>7. Case Studies in Scientific Methodology</h3><p>The question of whether there is a single scientific method is answered affirmatively. In other words, the motto holds: “There is One Method,” just as a chemist remarked in a somewhat critical oblivion that there is only one way of doing it in chemistry. The scientific method is an historically accumulating portfolio of techniques, behaviours, methodologies, paradigms, and world outlooks to generate knowledge or enlighten existence. Variably, it is by generations of physicists more and more distil into an ever-feasible distinctive method. Notably, in the inner shape of a linearised vector space of the scientific method, it is suffixed by three spanned intrinsic independent dimensions. For this reason, such a single method is expanded to a hexagon with equidistant purposed radii of gravity customarily centred as one of these dimensions. Besides, it is investigated how modelling, simulations, and computing entered the learning process of these dimensions.</p><p>Except in its vector space presentation, the scientific method was firstly naturalised in a rather peculiar, subjective, and professional sense: being a unique one. This unplanned procedure therefore amassed and composed a remarkable amount of knowledge encompassing causation principles, counter-examples, known-un-knowables, global behaviours, and local facts. As a result, 1 ± 0.5 scientific methods exist. Mathematical techniques have been axiomatized by theorists, such as checking preconditions, and rules of length, point, curve, formula, constant, function, letter, index, and proof. A procedure for science describes the use of inspecting a layout to derive the axioms for lattices, girders, sewages, and other man-made observables. Science is alp and omega, general and optimal, systematic and pragmatic, and direct and indirect to examine nature. It has various techniques to inspect chaos and turning points, singularities and catastrophes, orders and patterns, continuum and scaling, and thresholds and downsizes, as well as ontology, hystery, ethos, and mystique of quantum physics.</p><h3>7.1. Case Study 1: Physics</h3><p>Science has obvious practical applications. Yet the very success of science provides us with a problem. Why do we trust scientific theories? Science is undoubtedly a source of knowledge. It has brought about enormous technological advancements. In the last few centuries, science has demystified many natural phenomena and discovered hitherto unknown entities, events, and relations among them, such as radioactivity, germs, black holes, quarks, and genes. Science provides us with theories that successfully predict the motion of planets, calculate the propagation of waves, and explain the stability and chemical properties of atoms. Scientific theories have led to the invention of incredibly complex devices and machines, which probe ever more deeply into nature and provide insights into the workings of the world that were previously unimaginable. The practical applications of scientific theories range from the prosaic or everyday life, such as flight, electricity, and telecommunications, to the fundamental, such as the genesis of the universe and life itself.</p><p>Yet the more science is able to penetrate ever more deeply into the fundamental structure of the world, the more enigmatic it becomes. Our pictures of nature have shrunk tremendously. The aldosterone is now a point-like object seemingly described by elements of a non-commutative C*-algebra, running on a 10-dimensional non-commutative space, and particle-like entities were deemed an emergent feature. And the question remains: how could (an engendered) intelligence capable of scientific theorizing come about in such a world? The difficulty lies in the fact that with the sharpening of knowledge, a greater gulf appears between epistemology and what there is. One of the greatest challenges of philosophy is to reveal why and how trustworthy knowledge is possible at all 17.</p><h3>7.2. Case Study 2: Biology</h3><p>To investigate how birds learn to identify their chicks, we designed an active learning case study using the scientific process centered on a recent biology research article. The case study consists of four modules: 1) Introduction to similar research stories, hypotheses, and identifying variables. 2) Predicting the results of a published study and learning how to graph the results. 3) Using results from the published study to confirm or reject hypotheses. 4) Learning how research studies progress and the potential involvement of undergraduate researchers. This lesson was well received, with students expressing interest in undergraduate research and a better understanding of the experimental design. This active learning strategy could be adapted for similar aims and topics in other biology courses 18.</p><p>Coming up with ideas for future scientific research studies is one method for doing science. This requires knowledge of current scientific findings, but it can also be fun and interesting. Most students, however, have little experience with scientific research or knowledge of what makes questions scientifically testable. By establishing the framework of the scientific method for investigating a question proposed by authors and then going through this research with the class, students gain experience in how scientific research proceeds and what methods can be applied to their own questions 19. This case study describes a classroom scenario to teach research design. One lesson involving a peer-reviewed research article is detailed, along with additional suggestions to adapt it for other questions and courses. This lesson has been used successfully for undergraduate statistics courses (with a similar structure). Because many students have little to no experience with scientific research, biological statistics, or statistical software, this lesson introduces students to the general ideas and terminology behind these topics. The goal of knowing your own is to introduce topics that may otherwise seem inaccessible. Students use an established framework (the scientific method) and peer-reviewed paper as scaffolding to navigate the methodologies and terminology in a similar classroom setting. Adapting this lesson for advanced courses may not require as extensive an introduction and could instead provide the methodology and data for students to analyze on their own.</p><h3>7.3. Case Study 3: Social Sciences</h3><p>The following case study is a lesson aimed at educating students about the scientific method through the exploration of a scenario wherein birds must identify their own offspring among many chicks in a nest. From experimentation paradigms similar to this study’s design, a sizable body of research exists examining how many species of birds learn to recognize their own chicks. By first laying the groundwork of a research question, scientists can delve deeper into a particular topic. Within this study, the question addressed is how birds are able to identify their own chicks among many in a nest, despite the fact that these chicks appear similar both morphologically and vocally. Cognitive/ethological experiments utilizing four paradigms are designed to begin addressing this question 19. The first policy implies that birds must know their own. It can but does not need to take place in the absence of chicks. The second policy suggests that chicks must somehow alter their calls or appearances to indicate their own, and this would likely need to take place only during the early post-hatching period. The third policy suggests that parents must learn and remember their own chicks vocalizations and/or appearances. Fourthly, individuals may learn models with or without consensus. These models may include parents, play mates, other species chicks or adults, and other calls that are not those of chicks.</p><p>The goal of this lesson is to get students to consider the general principles of the scientific method for investigating phenomena in the natural world, as well as the particulars of studies regarding how birds learn to recognize their own offspring within this context. Small group discussion questions focused on the key steps of the scientific method are addressed as a class to ensure comprehension throughout the lesson. Additionally, students work in small groups to create graphical predictions that would support the hypothesis being tested within studies regarding this question, hopefully sparking their scientific creativity. Questions that prompt students to think about both what future work should be done to investigate a topic further, as well as how ethical issues pertaining to the scientific method could be discussed in a classroom environment are asked. Throughout the lesson, students are actively engaged in discussing answers to prompts and developing graphs themselves, thus using some approaches to learning principles.</p><h3>8. Challenges and Limitations of the Scientific Method</h3><p>In a world increasingly driven by technology, and in the life sciences in particular, the discussion of what constitutes valid experimentation requires reconsideration and formalism. The main idea is to take one step in this direction by broadening Karl Popper’s classical scientific method into a three-dimensional framework 1. In addition to its deductive dimension, two dimensions of inventive reasoning are introduced, namely, inductive and allochthonous reasoning. The argument is made that, although all three dimensions are at the basis of many successes of modern life sciences, they also each come with inherent challenges. A newly created metric provides the basis for assessing possible approaches as well as experiments. Much confusion about scientific validity arises from a fuzzy understanding of these different types of reasoning. It is hoped that this general framework will contribute to a better understanding of the nature of challenges being brought into research by the misuse of reasoning methods. Consideration of this broad type of scientific reasoning is not new, and — as will be shown — challenging aspects have been discussed previously, albeit more or less separately and without the benefit of a general framework. Copernicus paved the way for modern astronomy, and experimental invention shaped the early years of molecular biology. In recent decades, the rise of big-data generation and analysis, as well as mathematical modeling, changed the modus operandi of scientific discovery. Traditionally, the scientific method had learned to work inductively, by forgoing wild fishing expeditions, choosing a simple base upon which to build a model, and only testing one parameter at a time. Then scientific progress occurred through the preparation of an experiment, its careful execution, and detailed, cautious analysis of the data. Every parameter had to be well understood, as did the balance of the experiment. The goal was not simply to obtain a significant result; much more was at stake. A model was to be proposed as the connection between the observables and their source, which meant that the effective functional form needed careful justification. To the traditionalists’ horror, the absence of such a clean framework became common in life science disciplines. New algorithms for model fitting and parameter estimation wrote codes that were so complex that they had to be treated as black boxes. As a result, important insights and paradigm shifts became impossible.</p><h3>8.1. Limitations of Empirical Evidence</h3><p>Empirical evidence is often considered the gold standard of good measurement in testing out scientific theories. The process of being proven thus truly, it is believed, is the hard test that any ‘truth’ of this world must undergo. However, empirical evidence has limits of coverage in scientific truth. For instance, what results from experiments such as population, cast and cast extension remain adamantly non-empirical, even though these results together with each other call for measurement results on yet more empirical issues to corroborate population. There are, moreover, also limits of use of empirical evidence. For instance, even though we now know that the distribution of particle masses hold, because it meets density and distributional function results based on empirical evidence, the existence of a non-empirical infinity problem renders the best fit of the Standard Model necessarily wrong 20.</p><p>Manifold extensions of QFT testable at high energy densities and damping energy ejvj scale automatically follow. They each assume a range of perilous input values, assumptions, collateral theories, and border effects that govern measurement. Yet, synthetic results of S-matrix elements formed outside the bubble cannot be empirically tested, as it is believed that QFT splits beyond locality. A non-empirical state of affairs, one that cannot be tested via S-matrix elements according to the general bulk-boundary prescription in quantum physics, consequently follows that forecludes any measurements of the bubble exterior, even if the bubble interior can be empirically tested 21. It needs careful consideration of such limits of empirical truth — together with philosophical analysis on how such limits lead to other permissions and prohibitions on its exploration and preparation of analytic tests on better understanding nature — in future projects assigned on the Scientific Method.</p><h3>8.2. Human Error in Research</h3><p>Studies in human error should provide useful insights into spreadsheet research. They measure the frequency of human errors and human performance. Most of the substantial research in this domain is in writing research which has similar content. When people write, they think of the grammar of the text and what they are going to say. Writing is mentally demanding. The more complex the grammatical constructs or thoughts being expressed, the greater the error rates. Research has shown that human error rates are on a scale from 2% to 12% at the level of simple thoughts / phrases, 10% to 15% at the level of noisy text, 15% to 30% at the level of sentences, and 95% to 98% at the level of complex thoughts such as groups of sentences / paragraphs. Surprisingly, this is the common sense behind the acceptance that one must proofread every important document. It follows that any document worthy of publication and nontrivial length (longer than a couple of lines) will contain grammatical or spelling errors, that is, errors in what is written 22.</p><p>Some of the most serious writing errors are in word omissions, duplications, or substitutions that badly obscure the writer’s intention. Formula errors, such as missing formulas, badly incorrect formulas, and misplaced formulas, are likely editorial errors in spreadsheets. In both writing and spreadsheet usage, semantic errors seem to go undetected. These are misunderstandings or confusion about the underlying model. Writing and spreadsheets embody complex models. The more complex a model is the greater the probability of a misconception about it. The power of a spreadsheet gives it the capacity to express complex models and equations. Further, unlike programming errors, complex semantic errors are likely to escape detection because they can generate misleading results that appear plausible. Both extensive testing and screening of spreadsheets would not find any errors despite cumbersome models. So a long period in which “no errors were found” would simply mean that the model was so wildly wrong that the results were incomprehensible before the machine broke down. Hence spreadsheet design sophistication is one of the best predictors of error frequency.</p><h3>8.3. Funding and Resource Constraints</h3><p>One of the greatest problems confronting an investigator today is how to secure continued funding of a project. Obvious economic viability and present-day utility of results is an important consideration in obtaining funding. A funding agency’s problems are often quite different as its major goals are to maximize the benefits to society and to minimize the expenditures of resources. The sequential nature of research, development, or exploration requires optimal funding criteria that allow present funding decisions to take into account future funding decisions 23. Thus, there exists a continual possibility of making revisions in a project’s status based on the newest information. Because decisions on funding are generally based on careful examination of the potential benefits and costs involved in the alternate courses of a project, the detailed and mathematical nature of this examination will be treated more rigorously than has heretofore been common in the literature.</p><p>Projects generally consist of a fixed number of stages, during exactly one of which all the benefits may be received. Benefits are uncertain at each stage, and are not received until completion of the project. As stages are completed, additional information about benefits is received. The rules which follow specify optimal funding criterion based on this structure of projects. Consideration will be given to their effects on maximizing the present value of total benefits to society. Specific funding criteria which reflect the sequential nature of projects in the funding decisions of a funding agency are considered. The present distributions of benefits resulting from several special cases of these funding criteria, as well as alternatives to them, are compared via a numerical example 24.</p><h3>9. Future Directions in Scientific Methodology</h3><p>In modern societies, the scientific method continues to evolve and expand. Despite this, a standard definition remains desirable, and most importantly, it is feasible 1. This does not mean that there is just one scientific method. Just as there is no single algorithm guaranteed to solve all algorithms, there is also no single methodology suitable for all scientific tasks. Nevertheless, it is proposed that the scientific method be viewed as a five-dimensional vector space of methodologies, or the specification of a standard basis that can facilitate clearer communication.</p><p>Some methodologies are better suited to specific tasks, and while there have been long and heated arguments between supporters of various methodologies, there is a need to abandon judgment about what constitutes science or pseudoscience. The methodology of falsification is versatile and clean, but it cannot address exploratory and descriptive aspects, nor can it be reasonably applied to optimal control or pre-hoc simulation, predictions, or interpretations of effectors. A recent surge of methodologies has provided a rich arsenal. The focus now is less on whether they work than on how to understand them and apply them rationally.</p><p>These methods are also not without their problems. Nevertheless, language is insufficient for rational description, and because mathematics lacks suitable syntax, mechanisms to guide parameter selection, and procedures to relate coarse- to fine-scale representations, these methodologies tend to fall into epistemic black boxes. When they work, they do so from being enacted as logical elaborations of empirical generation of postulates through repeated questioning of a model by experimental data. When they do not work, it is unclear whether robustness patterns hold, with resulting large data sizes creating further concerns.</p><h3>9.1. Interdisciplinary Approaches</h3><p>Several disciplines have been discovered and developed, with their own principles. However, a focus on methods, algorithms, and epistemological principles common to many disciplines is revealing. Following a few examples in various disciplines, pedagogical traditions such as the scientific method, empirical science, the conjecture-and-refute principle, the hypothetico-deductive method, and the critical-experimental method are described 1. Algorithms common to many disciplines are then detailed, why they work, and how to modify them. Important aspects typically not addressed in scientific method textbooks are also highlighted, such as how to deal with the discovery and formulation of the question, the analysis and synthesis of the experimental concept, the analysis of robustness and uncertainty, the fitting of deterministic data, the use of multiobjective optimization for exploration and input design, and the development of robust signal-processing methods distinctive for nonlinear dynamics.</p><p>Analogous interdisciplinary approaches and paradigms have been proposed. The cross-disciplinary perspective is particularly needed for the AI effort, particularly to propose paradigms for problem formulation that have worked in a substantial number of disciplines and to reproduce human abilities in the same way. Nevertheless, it is doubted that simple principles, such as limited precision, will scrape the surface of abstracting many of the more capable abilities. This chapter is meant as an advancement toward a comparative scientific methodology that sheds light on methodology, reasons why certain approaches work, and understanding how to piece together problem formulation and solution, which is common across fields. Two examples are provided, one from medicine and bioengineering, as well as a short overview of another from micro-machining. Global applications with complexity attributes that vary enormously are also featured.</p><h3>9.2. Integration of Technology</h3><p>Science and technology are intertwined. The 21st century is an age of technology, but unlike earlier technological revolutions, the growth of technology has become exponential. The advancement of science and technology is breathtaking and arouses apprehensions about managing the invisible changes occurring ( 25 ). Science is based on empirical data and relies heavily on meticulous measurement. Technology, or technological artifact, is the mean by which data are collected. 3D spectacles are the technological artifacts with which students can virtually see how to build a reconciliation map. It is like having a curator on site for natural history exhibits. One does not merely glimpse the statues of the ancient Pharaohs. A slick movie, accompanied by carefully orchestrated sounds, sets the scene. The melting snow caps of Kilimanjaro tumble toward the sea with boulders. To capture student imagination and increase their view of the subject, immersion into the actual learning scenes is crucial. The information that 3D resources obtain virtually constancy accompanies enormous amounts of data collected by cameras, probe station devices, and many other innovative technos.</p><p>Without thoughtful and systematic consideration of the benefits/drawbacks, products might backfire and create learning barriers. The effort to examine the best classroom practice of using technology has cherished a decade of discussion. Science classrooms with resources become the focus of this study. The important view on the results of the study emphasizes that technology’s affordance and constraint in pedagogy can help to dislodge any preconceived idea of a new technological resource that might be an ultimate panacea. Webb’s epistemic criteria reveal that this case study in a rural high school rates the lesson in all but one of the levels as being midrange in uptake. Eve’s analysis of the classroom conversation offers compelling rationale for this dimension. Because of the format of the technology itself, the 1st two levels learning tasks were under-utilized and consequently less comprehension/understanding occurred. Improvements in both these regards suggest that the appropriate choice and design of technological resources may enhance comprehension and classroom discussion in the future. Many teachers use a plethora of commercial e-tools: they apply online magazines, weather satellite images, pedo-climatic maps, and visible earth pictures randomly. The integration of these resources becomes unprecedentedly fluid in the information age.</p><h3>9.3. Citizen Science and Public Engagement</h3><p>Citizen science is a key advancement in data collection approaches, particularly in the realm of so-called “big data”. Despite the technical difficulties in gathering large datasets, advancements in internet infrastructures, widespread use of mobile devices, and the integration of large datasets with predictive algorithms have made this possible in other fields. A few data-collecting community-based efforts have arisen due to this situation, aiming for a greater understanding of social effects and phenomena that could not be captured before either because of the required infrastructure, class and privacy issues, or simply the lack of volunteers. However, these ventures are mostly ad-hoc and in a single vernacular language, which greatly limits their scope. With the objective of experiencing a collective, gradual, and cross-disciplinary effort to adapt classical basic astrophysics experiments to citizen sciences, the PosAm experiment was designed 26.</p><p>The PosAm experiment is based on the gravitational lensing effect present in many astronomical observations. This effect causes images of an astronomical source to be distorted, duplicated, or triplicated depending on its relative position to the foreground mass. It is possible to parameterize and fit the projected mass distribution to the data, recovering the model parameters (specifically, the positions, velocity dispersion, and ellipticities). While this is typically done using high resolution images from telescopes, a simplified version of this effect occurs in fresh orange juice poured in a glass. In this type of experiment, objects inserted in the tank yield radial light stretching while the edge refracts light convergently through the creative manipulation of the refractive indices of the material 27. The responsibility of designing the experiment and programming the catching procedures is on scientists. Once set up, it is turned on and off and volunteers are required to obtain data. The analysis reveals the inner workings of the experiment and forces citizens to critically view their role in the data collection process. The gathering of positrons is a simple day-long task that entertains citizens, hence another interest arises. The analysis process enhances the understanding of uneven distributions in activities, such as when and where the gossip begins.</p><h3>10. Conclusion</h3><p>In conclusion, the scientific method is a multifaceted and ever-evolving approach to acquiring knowledge about the natural world. The roots of the scientific method can be traced back to ancient civilizations, but it has undergone significant refinement over time, particularly since the time of contemporary scientists. In exploring the scientific method, several key points have come to light, including the importance of clearly formulating research goals that address specific open questions, the need for mathematical-computational algorithms to derive models from observations or experimental data, and the role of experimental methods and strategy in assessing model generality, applicability, and validity. The scientific method can be stratified into three dimensions: deductive, inductive, and allochthonous. Each of these dimensions has advantages and drawbacks, and all are essential for scientific progress. Whether a mathematician wishes to solve a problem from basic principles analytically, a natural scientist develops a mathematical description of a natural system from observations through curve fitting, or a data scientist uses a vast array of procedures to discover patterns in previously unregulated data, all approaches can be broadly labeled as scientific. However, many difficulties, some also specific to a dimension, lie ahead. Therefore, the endeavor to better understand the different dimensions of the scientific method and to refine the corresponding operational tools should continue. Hopefully, this will enable scientists to do a better job when exploring nature and, thereby, to open new windows of understanding into the way life works 1. If the scientific method is perceived as a vector space with basis vectors corresponding to the presented dimensions, some progress beyond the status quo is anticipated.</p><p>Looking into the near future, obtaining large quantities of data and using diverse strategies to analyze them will likely continue to gain in importance, because neither the blurring of these methodological spheres nor an understanding of the many-layered resulting intricacies is in sight. The use of computational techniques is expected to increasingly penetrate all three dimensions of the scientific method, but the difficulty of obtaining usable semantics and its algorithmic complexity will march along. More broadly, the importance of a sound methodology and of transparent work together with a cautious interpretation of results must be reinforced 8. Whether contemporary problems with the credibility of scientific information can be mitigated this way remains an open question.</p><p><strong>References:</strong></p><p>O. Voit E. Perspective: Dimensions of the scientific method. Published 2019. <a href="https://www.ncbi.nlm.nih.gov/pmc/articles/PMC6742218/">ncbi.nlm.nih.gov</a></p><p>Lawrence Randall D. Formal Methods in the Foundations of Science. Published 1970. <a href="https://core.ac.uk/download/216151869.pdf">[PDF]</a></p><p>D’Agostini G. From Observations to Hypotheses: Probabilistic Reasoning Versus Falsificationism and its Statistical Variations. Published 2004. <a href="https://arxiv.org/pdf/physics/0412148">[PDF]</a></p><p>Brüssow H. On the role of hypotheses in science. Published 2022. <a href="https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9618321/">ncbi.nlm.nih.gov</a></p><p>Parke E. Experiments, Simulations, and Lessons from Experimental Evolution. Published 2015. <a href="https://core.ac.uk/download/214153768.pdf">[PDF]</a></p><p>Loukas O, Ryun Chung H. Total Empiricism: Learning from Data. Published 2023. <a href="https://arxiv.org/pdf/2311.08315">[PDF]</a></p><p>Brittain J, Cendon M, Nizzi J, Pleis J. Data Scientist’s Analysis Toolbox: Comparison of Python, R, and SAS Performance. Published 2018. <a href="https://core.ac.uk/download/216907952.pdf">[PDF]</a></p><p>Crupi V, Hartmann S. Formal and Empirical Methods in Philosophy of Science. Published 2009. <a href="https://core.ac.uk/download/148348556.pdf">[PDF]</a></p><p>Queirós A, Faria D, Almeida F. STRENGTHS AND LIMITATIONS OF QUALITATIVE AND QUANTITATIVE RESEARCH METHODS. Published 2017. <a href="https://core.ac.uk/download/236102108.pdf">[PDF]</a></p><p>Sullivan S, Brockington D. Qualitative methods in globalisation studies : or, saying something about the world without counting or inventing it. Published 2004. <a href="https://core.ac.uk/download/47618.pdf">[PDF]</a></p><p>Almeida F. STRATEGIES TO PERFORM A MIXED METHODS STUDY. Published 2018. <a href="https://core.ac.uk/download/236101754.pdf">[PDF]</a></p><p>Mendling J, Leopold H, Meyerhenke H, Depaire B. Methodology of Algorithm Engineering. Published 2023. <a href="https://arxiv.org/pdf/2310.18979">[PDF]</a></p><p>Mendling J, Depaire B, Leopold H. Theory and Practice of Algorithm Engineering. Published 2021. <a href="https://arxiv.org/pdf/2107.10675">[PDF]</a></p><p>Celiktutan B, Cadario R, K. Morewedge C. People see more of their biases in algorithms. Published 2024. <a href="https://www.ncbi.nlm.nih.gov/pmc/articles/PMC11032492/">ncbi.nlm.nih.gov</a></p><p>Felderer M, Horta Travassos G. The Evolution of Empirical Methods in Software Engineering. Published 2019. <a href="https://arxiv.org/pdf/1912.11512">[PDF]</a></p><p>Saint-Mont U. On the Logic (plus some history and philosophy) of Statistical Tests and Scientific Investigation. Published 2018. <a href="https://arxiv.org/pdf/1801.01829">[PDF]</a></p><p>Dardashti R, Hartmann S. Assessing Scientific Theories: The Bayesian Approach. Published 2019. <a href="https://core.ac.uk/download/196238171.pdf">[PDF]</a></p><p>K. Hubbard J, Shizuka D, A. Couch B. Knowing your own: A classroom case study using the scientific method to investigate how birds learn to recognize their offspring. Published 2017. <a href="https://core.ac.uk/download/188134615.pdf">[PDF]</a></p><p>K. Hubbard J, Shizuka D, A. Couch B. Knowing your own: A classroom case study using the scientific method to investigate how birds learn to recognize their offspring. Published 2017. <a href="https://core.ac.uk/download/189481473.pdf">[PDF]</a></p><p>J. Oldehinkel A. Editorial: Factualities — establishing empirical truths in child psychology and psychiatry. Published 2021. <a href="https://www.ncbi.nlm.nih.gov/pmc/articles/PMC9293297/">ncbi.nlm.nih.gov</a></p><p>Barwich AS. How to be rational about empirical success in ongoing science: The case of the quantum nose and its critics. Published 2018. <a href="https://core.ac.uk/download/160114074.pdf">[PDF]</a></p><p>R. Panko R. Thinking is Bad: Implications of Human Error Research for Spreadsheet Research and Practice. Published 2008. <a href="https://arxiv.org/pdf/0801.3114">[PDF]</a></p><p>W. S. Roberts K, Lawrence Weitzman M. Funding criteria for research, development, and exploration projects. Published 1979. <a href="https://core.ac.uk/download/4397861.pdf">[PDF]</a></p><p>R Dougherty E, Shmulevich I. On the Limitations of Biological Knowledge. Published 2012. <a href="https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3468890/">ncbi.nlm.nih.gov</a></p><p>M. Harradine T. Rationale for and Examples of Internet-Based, Software, and Hardware Technologies That Can Be Integrated into the Science Classroom. Published 2007. <a href="https://core.ac.uk/download/233570956.pdf">[PDF]</a></p><p>Garbarino J, E. Mason C. The Power of Engaging Citizen Scientists for Scientific Progress. Published 2016. <a href="https://www.ncbi.nlm.nih.gov/pmc/articles/PMC4798819/">ncbi.nlm.nih.gov</a></p><p>Sagarra O, Gutiérrez-Roig M, Bonhoure I, Perelló J. Citizen science practices for computational social science research : the conceptualization of pop-up experiments. Published 1970. <a href="https://core.ac.uk/download/132201922.pdf">[PDF]</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=947109b4913a" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Intellectual Property Management in DeSci Frameworks]]></title>
            <link>https://medium.com/@0x_Rorschach/intellectual-property-management-in-desci-frameworks-83b9d5649948?source=rss-0b668e3fef6c------2</link>
            <guid isPermaLink="false">https://medium.com/p/83b9d5649948</guid>
            <category><![CDATA[desci]]></category>
            <category><![CDATA[blockchain]]></category>
            <category><![CDATA[intellectual-property]]></category>
            <dc:creator><![CDATA[Rorschach]]></dc:creator>
            <pubDate>Sat, 30 Nov 2024 10:13:32 GMT</pubDate>
            <atom:updated>2024-11-30T10:21:39.434Z</atom:updated>
            <content:encoded><![CDATA[<p>The integration of blockchain technology into the management of intellectual property (IP) rights within decentralized scientific (DeSci) frameworks represents a transformative opportunity to improve collaboration, transparency, and security in research projects. As research becomes more collaborative and decentralized, traditional ways of managing intellectual property (IP) face big problems. For example, it&#39;s challenging to figure out who owns something and how to protect rights in a digital setting. Blockchain technology, with its inherent characteristics of decentralization, immutability, and transparency, offers innovative solutions to these challenges.</p><p>One of the key benefits of blockchain in IP management is its ability to provide a secure and tamper-proof record of ownership and transactions. This is particularly important in collaborative research environments, where multiple parties contribute to the creation of intellectual property. By using blockchain, researchers can create a clear and verifiable record of ownership that is accessible to all parties involved in a project. This transparency helps mitigate disputes over authorship and rights and creates a more collaborative atmosphere. For instance, Pang et al.&#39;s work underscores the necessity of robust IP management systems, emphasizing the significance of registering, confirming, and monitoring IP rights, a task that blockchain solutions can effectively tackle (1).</p><p>In addition, blockchain technology improves the efficiency of IP transactions. Traditional methods often involve cumbersome processes that can delay the registration and transfer of rights. Blockchain streamlines these processes by enabling real-time updates and automated transactions through smart contracts. Self-executing agreements, known as smart contracts, incorporate the terms directly into the code, enabling the automatic enforcement of IP rights upon meeting certain conditions. Ito and O’Dair discuss the potential of smart contracts in IP management, noting that while they may replicate some negative aspects of traditional systems, their benefits in terms of efficiency and security are significant (2).</p><p>The application of blockchain in IP management also addresses cybersecurity concerns. As research data becomes increasingly digital, the risk of unauthorised access and data breaches increases. Blockchain’s decentralised nature makes it inherently more secure against such threats, as it eliminates single points of failure that are common in centralised systems. TAK highlights the potential for blockchain to enhance cybersecurity across a range of applications, including IP management, by improving record-keeping and reducing access management costs (3). Collaborative research environments particularly require the protection of sensitive data and intellectual property from unauthorised access.</p><p>Furthermore, the use of blockchain can facilitate the establishment of a more equitable IP management framework. Traditional IP systems often favour established companies, leaving individual creators and smaller organisations at a disadvantage. Blockchain democratises access to IP management tools, allowing a wider range of stakeholders to participate in the protection and monetisation of their intellectual property. Nyame et al. emphasise the importance of effective knowledge sharing in a collaborative environment and suggest that blockchain can enhance this by providing a secure and transparent mechanism for IP management (4). This shift towards inclusivity can lead to a more diverse and innovative research ecosystems .</p><p>In addition to these benefits, blockchain technology can also support the development of new business models for IP commercialisation. By enabling direct transactions between creators and consumers, blockchain can reduce reliance on intermediaries, thereby increasing financial returns to creators. This is particularly relevant in the context of digital assets, where traditional licensing models can be cumbersome and inefficient. In their paper, Guo et al. suggest a blockchain-based transaction sharing system that allows IP transactions and protects privacy through proxy re-encryption. This system solves problems with both security and efficiency (5). This model can enable researchers to monetise their innovations more effectively, thus fostering a culture of innovation and collaboration.</p><p>However, the implementation of blockchain for IP management is not without its challenges. The legal and regulatory framework surrounding blockchain technology is still evolving, and there are significant hurdles to overcome in terms of standardisation and interoperability. As noted by Alqarni et al., while blockchain holds promise for improving IP management, the complexity of legal recognition and enforcement of blockchain-based IP rights remains a significant barrier to widespread adoption (6). In addition, the potential for smart contracts to inadvertently replicate existing legal issues requires careful consideration and design to ensure compliance with existing laws and regulations.</p><p>We need to address the issue of privacy in blockchain applications. While the transparency of blockchain is a key feature, it can also lead to concerns about the exposure of sensitive information. Yuan discusses the need for privacy-preserving methods in blockchain-based IP authentication, emphasising that while blockchain can secure IP rights, it must also protect the privacy of creators(7). This balance between transparency and privacy is crucial for fostering trust among stakeholders in decentralised research environments.</p><p>In conclusion, the integration of blockchain technology into the management of IP rights in decentralised scientific frameworks offers a promising way to improve collaboration, security, and efficiency in research projects. By providing secure, transparent, and efficient mechanisms for IP management, blockchain can address many of the challenges faced by traditional systems. However, careful consideration of legal, regulatory, and privacy issues is essential to fully realizing blockchain’s potential in this context. As the research landscape continues to evolve, the adoption of blockchain technology could play a pivotal role in shaping the future of IP management.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*LL7sqEc_va8LpL9kpUIVcQ.png" /></figure><p><strong>References:</strong><br>1. Pang Z., Yi W., &amp; Zhou H.. Research and analysis on intellectual property information management in military academies based on blockchain technology. 2023. <a href="https://doi.org/10.4108/eai.9-12-2022.2327731">https://doi.org/10.4108/eai.9-12-2022.2327731</a><br>2. Ito K. and O’Dair M.. A critical review of the application of blockchain technology to intellectual property management. 2018:317–335. <a href="https://doi.org/10.1007/978-3-319-99058-3_12">https://doi.org/10.1007/978-3-319-99058-3_12</a><br>3. Tak P.. The critical determinants of the application of blockchain technology in enhancing cybersecurity in the modern technology era. cienc.eng. 2023;11(1). <a href="https://doi.org/10.52783/cienceng.v11i1.214">https://doi.org/10.52783/cienceng.v11i1.214</a><br>4. Nyame G., Qin Z., Agyekum K., &amp; Sifah E.. An ECDSA approach to access control in knowledge management systems using blockchain. Information 2020;11(2):111. <a href="https://doi.org/10.3390/info11020111">https://doi.org/10.3390/info11020111</a><br>5. Guo Z., Su Y., &amp; Hu X.. Intellectual property transaction sharing scheme based on blockchain and proxy re-encryption. 2023. <a href="https://doi.org/10.1117/12.2678856">https://doi.org/10.1117/12.2678856</a><br>6. Alqarni M., Alkatheiri M., Chauhdary S., &amp; Saleem S.. Use of blockchain-based smart contracts in logistics and supply chains. Electronics 2023;12(6):1340. <a href="https://doi.org/10.3390/electronics12061340">https://doi.org/10.3390/electronics12061340</a><br>7. Yuan S.. A blockchain-based privacy preserving authentication method for intellectual property. Symmetry 2024;16(5):622. <a href="https://doi.org/10.3390/sym16050622">https://doi.org/10.3390/sym16050622</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=83b9d5649948" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Exploring the Concepts and Applications of Boolean Satisfiability]]></title>
            <link>https://medium.com/@0x_Rorschach/exploring-the-concepts-and-applications-of-boolean-satisfiability-90d255a81878?source=rss-0b668e3fef6c------2</link>
            <guid isPermaLink="false">https://medium.com/p/90d255a81878</guid>
            <category><![CDATA[algorithms]]></category>
            <category><![CDATA[boolean-satisfiability]]></category>
            <category><![CDATA[innovation-games]]></category>
            <dc:creator><![CDATA[Rorschach]]></dc:creator>
            <pubDate>Wed, 18 Sep 2024 08:37:49 GMT</pubDate>
            <atom:updated>2024-09-18T08:37:49.895Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/984/1*6gKyKsoKDtKB_DpOtW-_mg.png" /></figure><h3>1. Introduction to Boolean Satisfiability</h3><p>Boolean satisfiability is a fundamental problem in computational logic. In Boolean logic, the value of a variable is restricted to being a member of the set {0, 1}. These elements can intuitively be interpreted as the English words “false” and “true,” respectively. Therefore, a simple variable is either false or true. More complex propositions can be developed in Boolean logic with the use of the connectives AND, OR, and NOT, which can be combined to form compound variables or statements.</p><p>Therefore, a complex variable may be defined as x1 AND x2 (or both x1 and x2 are true), or as x1 OR x2 (or either x1 or x2 is true), or as NOT x1 (or x1 is false).</p><p>Lets Make it easy to understand:</p><blockquote>SAT is like solving a puzzle where you have to figure out if there’s a way to make a complicated rule true by choosing “yes” or “no” answers.</blockquote><blockquote>Imagine you have a bunch of light switches, and some rules say things like:</blockquote><blockquote>- “If switch A is on OR switch B is off, then the light turns on.”<br>- “If switch C is off AND switch D is on, then the light turns off.”</blockquote><blockquote>Your job is to figure out if you can set the switches (either on or off) in a way that makes all the rules work together. If you can do that, then the puzzle can be solved (it’s “satisfiable”). If there’s no way to make all the rules work, the puzzle can’t be solved (it’s “unsatisfiable”).</blockquote><blockquote>So, SAT is just about figuring out if there’s a way to turn the switches on and off to make everything work as expected.</blockquote><p>Computationally, the satisfiability of a logical statement (i.e., a combination of simple and complex variables connected by connectives) may be established using algorithms. These algorithms allow practical applications in computer science and computational logic to be developed. A few applications of Boolean satisfiability problems are found in the field of circuit design, graphical models, program analysis, security, and artificial intelligence, among others.</p><p>There are many scenarios where establishing the satisfiability of a Boolean expression is also important: a disguise of truth (i.e., tautology) and whether a statement is consistent with another statement or set of statements (i.e., satisfiable or contradictory). Efficient computational methods for generating the satisfiability of complex variables are based on their familiarity and result in the truth value of the logical statement, determining whether the statement is true or false.</p><h3>1.1. Definition and Basics</h3><p>Boolean satisfiability, or SAT, is the satisfiability problem that only applies to propositional logic. It is the task of figuring out if a given expression in a Boolean algebra (i.e., a Boolean formula) can be made true by giving the variables in the expression truth values. A Boolean formula is satisfiable if there exists some binary vector of truth values for each of the variables in the formula such that the expression evaluates to true.</p><p>In contrast, a formula is unsatisfiable if there is no possible combination of truth values for the variables such that the expression evaluates to true. In terms of computation theory, SAT is often cited as “the canonical NP-complete problem”; in other words, Boolean satisfiability problems lie at the heart of NP-completeness theory. The propositional satisfiability problem has also been the subject of a significant amount of scrutiny in the field of formal logic, especially as it pertains to complexity classes.</p><p>Propositional logic is one of the primary tools in finite model theory, and early research in this area has led to a number of important practical and theoretical results. Part of the utility of SAT and the Boolean satisfiability problem lies in the inherent determinism of propositional logic and its relatively straightforward handling of time. For readers familiar with SAT and propositional logic, this section can be safely skipped with no significant loss of continuity. For readers who are new to the area, basic definitions and algorithms will be defined below. We begin with definitions of basic concepts in SAT (Boolean expression manipulation is a prerequisite to discussing major improvements in modern systems).</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/460/0*wpI8vJKY9XuomHgi.png" /><figcaption>Knowledge Representation</figcaption></figure><h3>1.2. Historical Background</h3><p>Current research problems related to Boolean Satisfiability (SAT) are not necessarily intuitive; they are influenced by historical inquiries deeply. Therefore, historically, it is possible to turn back the days to begin investigating SAT. The theory of logic and computability has a strong connection with the mathematical subfield of “Boolean logic,” which received its name from the English mathematician and logician.</p><p>In computer science, logical questions are even more vital: for all practical purposes, the yes/no questions. The question of deciding whether formulas in classical propositional logic have an “interpretation” that makes the formulas true is reducible to the question of whether graphs possess a certain property. When logics have more structure and more expressive power than propositional logic, such as the predicate logic of the previous example, logical problems become more difficult.</p><p>Despite this, if we reintroduce the question of asking whether or not any “<strong>interpretation</strong>” would satisfy a formula but this time restrict attention to formulas with specific constraints on the satisfaction of no repeatable variable, this search gets easier. This is the core SAT problem. Moreover, despite the significance of SAT as an autonomous research issue, it is of deep interest to these inquiries because of its weight.</p><h3>2. Algorithms for Solving Boolean Satisfiability Problems</h3><p>Various algorithms have been proposed to process and solve Boolean satisfiability problems systematically. Such tools are crucial in making automated systems more intelligent by providing them with adequate computational decision-making capabilities. This section sheds light on some well-known algorithms developed to solve SAT problems. I am going to begin with the DPLL algorithm, which is considered the fundamental source for various modern SAT-solving methodologies. The transition from the DPLL algorithm to Conflict-Driven Clause Learning methods has introduced more advancements in the state of the art by keeping the spirit of computing longest lemmas. We touch upon the general principles behind the developed algorithm, which will help to explain their essential paradigms.</p><p>The major advantage of this approach is that a better understanding of the classic DPLL procedure, as a fundamental frame, is helpful to illustrate the current state of the art SAT-solving techniques effectively. The theory and practice explained on the DPLL-enhanced procedures can shed some light on the progress of the recent SAT solvers. The procedure marked the first significant attempt to solve SAT problems systematically. I will take a general look at the extended DPLL approaches before heading to the newly developed CDCL class, which originated from the DPLL base.</p><p>This is the current state of the art in SAT solving and is referred to as the CDCL category algorithm. In order to compare CDCL with the classical DPLL methods in terms of SAT solving performance, the basic characteristics of the DPLL methods are compared with those of the CDCL approach. In the SAT-solving domain, numerous CDCL solvers have been successfully introduced, either when competing with each other in various benchmark instances or when applied to practical problems.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/781/0*31ZYfUJRQn_OTYw2.png" /></figure><h3>2.1. Davis-Putnam-Logemann-Loveland (DPLL) Algorithm</h3><p>The problem of deciding whether there exists an assignment that satisfies a given propositional logic formula, commonly referred to as a SAT problem, has significant practical and theoretical implications in the areas of automated reasoning, formal methods, and computer-aided design, among others. One of the most successful methods for solving these types of problems has been the DPLL algorithm. The DPLL algorithm introduced a systematic recursive strategy for rewriting and simplifying propositional logic formulas based on certain strategies and properties. This algorithm established some of the fundamental concepts that constitute modern SAT solving and helped popularize the study of SAT solvers from a theoretical and practical perspective.</p><p>DPLL exploits the notion of unit clauses, which allow the algorithm to systematically simplify the input propositional logic formula through a process called unit propagation. The effectiveness of DPLL, however, comes at the expense of the time and memory costs of executing the algorithm in practice. Providing an in-depth overview of DPLL and its analysis follows. In some cases, the DPLL’s search technique has been shown to be more efficient and effective than other standard search techniques, particularly when combined with modern heuristic functions. Various case studies from a variety of domains effectively utilizing DPLL are also introduced.</p><h3>2.2. Conflict-Driven Clause Learning (CDCL)</h3><p>The development of the conflict-driven clause learning (CDCL) algorithm is regarded as a milestone in the world of SAT solving. CDCL is a huge leap forward from other DPLL-based complete procedures and has triggered exponential progress in this domain. It is the earliest published algorithm of a modern SAT solver that uses the latest technologies thoroughly and is capable of solving industrial SAT instances. The crux of CDCL is to perform conflict analysis, to learn new clauses, and then to return to the search until it can find a solution. The abilities of conflict analysis and clause learning also make CDCL particularly efficient in solving large and complex CNF formulas. In summary, a CDCL solver is generally compared favorably to a modern DPLL-based SAT solver in terms of some important characteristics such as solving industrial instances, adaptability, and learning-aided solving.</p><p>Refinements of the conflict analysis and clause learning strategies lead to the ability of today’s CDCL solvers to solve industrial instances with up to millions of variables. The power of the CDCL approach is clear in many fields where NP-complete problems’ satisfiability tests are often required. During that time, polynomial-time domain-specific algorithms may not be available. The CDCL approach far exceeds the DPLL framework algorithm in practical application software.</p><p>The most advanced solvers are used efficiently in hardware verification, AI, planning, cryptography verification, and many other fields that rely on SAT solvers.The applications of modern CDCL SAT solvers are broad and important in the real world. The impact of CDCL SAT solvers on satisfactory and complete problem research is unparalleled. Many researchers in academia and industry are working on the development of new algorithms and the utilization of the latest hardware for solving huge instances.</p><h3>3. Practical Applications of Boolean Satisfiability</h3><p>Because SAT solvers are increasingly powerful, reducing hardware problems to SAT problem solving and software problems to SAT problem solving is becoming the most common tool in academia and industry for ensuring the correctness of digital systems. In general, the problem is huge. There could be as many as 250,000 to 500,000 gates. To show that the system is built correctly, a designer must demonstrate that the system is contained in a system with 250,000 gates. Here, formal methods are used. When employing the satisfiability concept in AI and machine learning, various intelligent systems have been studied that focus on decision making with combinatorial optimization as an important aspect.</p><p>Some of these systems include diagnostic systems, identification systems, intelligent manageability systems, design verification systems, and agents. Some examples of intelligent systems are a diagnostic system that uses exhaustively calculated values and rules to confirm the candidate with a constraint optimization problem that is used to minimize the size of the candidate. Since none of the states or atoms of the candidate violate the constraint, no operator should remove any states or atoms of the candidate; this is satisfied. The next subsection provides more detailed examples.</p><p>The use of modern solvers is possible in these applications. For example, in a diagnostic system, the complementary binary decision diagram is utilized. To ensure the actual executability of the solution, the declarative solution is evaluated through counterexample checking. These pragmatic examples above show that Boolean satisfiability seems to be slowly seeping into commercial ventures.</p><h3>3.1. Hardware and Software Verification</h3><p>Reliability and correctness have always been the major concerns for both hardware and software in computer science. In the last two decades, researchers have made a lot of efforts in exploring various techniques and methodologies for verifying and validating the correctness of hardware and software systems. The increasing size of the combinatorial explosion of modern programs and microprocessors and their underlying components makes the development of automated verification and validation techniques more challenging. To address this issue, researchers have been exploring the Boolean satisfiability problem, which has been proved to be NP-complete. Research efforts in Boolean satisfiability have paid off recently, and a number of efficient SAT solvers are available in the public domain and many tools development.</p><p>A significant amount of silicon design effort is invested in the development of an efficient front-end verification process to ensure the correctness of digital designs. The industry follows rigorous design methodologies and exhaustive simulation-based verification techniques to confirm the various requirements of a design. As an alternative or additional step to this costly simulation-oriented approach, formal verification techniques based on model checking and theorem proving have been developed. Another promising technology that can make formal methods more widely available is Boolean satisfiability.</p><p>The design of efficient SAT solver technology has led to some recent success in hardware verification. Several significant case studies in applying SAT technology to hardware verification have been published. This work is impacting the industry and increasing the confidence that SAT-based techniques are an important part of future computer-aided formal verification and simulation tools. Software developers are striving to ensure the reliability and robustness of their software applications in the fast-growing internet and mobile device era.</p><p>Moreover, software developers use various quality assurance tools, including testing, equivalence checking, and model checking tools, to ensure software quality. One of the successful methodologies for verifying the reachability of safety properties and discovering errors in software applications is model checking. SAT was used as a backend solver in software model checking. The development of efficient practical algorithms for deciding the satisfiability of propositional formulas has a huge practical impact on improving the performance and existing capacities of many tools and techniques.</p><h3>3.2. AI and Machine Learning</h3><p>AI and machine learning environments have been one of the most influential contexts for the integration of SAT concepts. At an operational level, Boolean satisfiability is important in the optimization tasks that use decision procedures such as SAT solvers. Research in KRR incorporates both the development of techniques for efficient representation and inference upon entities reducible to logic and the study of representation, reasoning, and negotiation with logic-like frameworks which allow large amounts of uncertainty. Graph properties and systems of equations are examples of artificial intelligence concepts which are then cast in logic and amenable to treatment by a family of SAT-like techniques. Solvers for text are beginning to contribute conceptual factors as interesting comparators. Many AI techniques are interested in the relationship between instances of agent-relevant state.</p><p>Problem constraints typically express state properties that can be determined by a representation, such as “looks good to A” =&gt; “looks good to B” or “agent A attached to site A can transport any liquid from agent B to site A”. The inferred property, such as goodlooks(A, Y), can then be added to the representation. An arbitrary property P(X,Y) can be specified by rules such as choose X, choose Y: such that Q(X,Y), assign to Z: Z = c and use the SAT solver to test if some instance of P exists.</p><p>This representation can be derived from a weak theory of arbitrary complex structural constraints. Other AI-related work has failed because tight model assumptions may not hold in cases of complexity. Static Boolean logic is more general and simple, and the SAT works better for complex topological quantitative models than modal logic as it uses cheaper quantifiers and disambiguation information.</p><h3>4. Current Research and Future Directions</h3><p>Despite remarkable achievements, the research landscape in the SAT domain continues to be one of active exploration. Recent editions have hosted a large number of SAT-related events, highlighting the high relevance of our field. Interdisciplinary collaborations are emerging across different areas, resulting in new challenges, methodologies, and knowledge transfer. Overall, challenging application domains such as DNA placement have gained widespread attention, driving research on ordered libraries.</p><p>The development of new methodologies combines advances in state-of-the-art algorithms with new theory and applications. It can be expected that the combination of both will be advantageous in the future. The new advances in resolution disproofs, with new solvers, tools, and techniques represent major breakthroughs that provide promising means to tackle the problems of unusual applications and theory, the growth in solving complex problems via algorithm selection and parameter configuration, efficient and scalable multi-core parallel SAT solving, while also providing new aspects in SE and MPS. During the Industrial Track, research collaboration opportunities on benchmarks for very satisfactory problem solving have emerged, and multiple solvers have been ported to different embedded devices.</p><p>Challenges in applied industrial interconnected circuit systems are a hot topic within the Industrial Track. New and dynamically adapting methodologies have emerged to solve challenging industrial power estimation and analysis problems in SAT. In the parallel and distributed constraint solving domain, two new exciting research directions are emerging. The first is in the area of application of parallel and distributed SAT solving in grid systems, while the second is to develop parallel multi-agent systems to solve hard-to-solve real-life constraint satisfaction problems. In the field of AI and the web, a new trend of developing an online SAT discussion community was initiated.</p><p>One of the society’s goals, in addition to delivering research, is a focus on education. The competition and evaluations are being used in complex training settings. In addition, using SAT domain knowledge in other knowledge representation fields shows applicability to a broader field than initially considered. With the strength of commercial SAT solvers ever increasing, it is necessary to consider larger and more exact benchmarks for the evaluation. Currently, non-commercial tool solvers that have been evaluated tend to either time out or produce incomplete results given an appropriate time frame.</p><p>Many interesting developments in theory have occurred over the previous contests since 2002. The category of comparisons and controversies are interesting new benchmarks that exhibit great potential. The hardware verification domains still require non-trivial adaptations to run under the new evaluations. The competition is healthy, but a substantial drive is necessary to make upsets occur in the graph theories domain. The growth of DNA-related type 2 research areas is always inspirational to follow. In general, we can see research adaptation to computational restraints of the present day. Currently, the challenge is able to place current state SAT research. We assume our current casting is indicative of the newly forming technology of SAT research and where it is heading. The challenges reported in our previous section are devoid of any write-up entries to date. We are working with the challenge entrants to create a fuller response for publication.</p><h3>5. Conclusion</h3><p>Here, a few of those points are collected into a few concluding remarks: Algorithms are the most significant tools for solving satisfiability problems, and the time and space complexities of these algorithms provide important insights into the problems. <br>(Thats why Innovation game is really matters)</p><p>The advances in the techniques and improvements in implementations have real-world consequences. The technologies depend on challenging computational problems that are efficiently solved by SAT solving algorithms, and the resulting implications and applications are huge. For these reasons, future work in this field centers on these algorithms, from other applications in distributed computing to problem structures in constraint satisfaction problems.</p><p>A foundational symbiosis between principle and application early and throughout all of these sections implies that such work will be essential to the future of research in this area. Advancements in the practical front (especially large-scale applications such as machine learning) are likely to define the direction of SAT research in the future as they have for the past 20 years. Furthermore, as each new paper from all four sections suggests, the field is full of interesting research opportunities that could impact other fields as much as, or even more than, satisfiability research impacts this field.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=90d255a81878" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Programmable Privacy and On-chain Compliance]]></title>
            <link>https://medium.com/@0x_Rorschach/programmable-privacy-and-on-chain-compliance-c4576c681dec?source=rss-0b668e3fef6c------2</link>
            <guid isPermaLink="false">https://medium.com/p/c4576c681dec</guid>
            <category><![CDATA[on-chain-compliance]]></category>
            <category><![CDATA[privacy]]></category>
            <category><![CDATA[blockchain]]></category>
            <dc:creator><![CDATA[Rorschach]]></dc:creator>
            <pubDate>Wed, 11 Sep 2024 08:48:59 GMT</pubDate>
            <atom:updated>2024-09-11T08:48:59.985Z</atom:updated>
            <content:encoded><![CDATA[<blockquote>Privacy isn’t about hiding; it’s about controlling what you share. It’s like having a door instead of a wall-it lets you decide when to open up or keep things to yourself.</blockquote><h3>Programmable Privacy and On-chain Compliance</h3><h3>Introduction</h3><p>Privacy is a fundamental building block of society. It lets us solve problems involving verification, safety, and security by allowing us to share sensitive or personal information in a controlled manner. In our increasingly digital and automated world, privacy is also a necessary component of digital transactions. For many, being able to efficiently prove compliance with regulatory requirements in a privacy-preserving fashion is increasingly a pressing concern. Businesses and private individuals are looking for mechanisms and solutions that help verify and prove compliance with a wide range of regulations on the one hand and privacy-compatible handling of know your transaction requirements with an eye on forthcoming guidance on the other.</p><p>Privacy-enabled smart contracts enable businesses to benefit from blockchain technology more fully. They will empower developers to create new privacy-preserving products and services in general and, specifically, will open up new opportunities for privacy regulators to facilitate the development of innovative solutions that simultaneously solve for privacy and compliance. I explain the fundamental concepts surrounding privacy-enabling blockchains and the challenges many of the technologies grapple with. I describe the need for effective privacy solutions, an issue that becomes all the more pressing as more people adopt blockchain solutions. As a result, privacy-enabling blockchains and sidechains are now being productionized and integrated with real-world applications.</p><p>I describe programmable privacy and explain how the programmable properties of blockchains offer a much stronger nexus between compliance and privacy. That, in turn, can open the door to new privacy-preserving applications. I provide a few real-world examples of how firms are already using programmable privacy to ensure data integrity. I provide a more detailed technical exploration of the constraints around enterprise data privacy by introducing the novel concept of stateful privacy accounts. I then sketch a vision for how such an account could be used.</p><h3>1.1. Background and Motivation</h3><p>In the digital age, privacy has been the subject of intense public scrutiny and academic inquiry. Going back to the early days of the web, it has been clear that extensive data collection and the loss of privacy posed threats to individuals and organizations. Two decades on, this concern has not diminished. In recent years, large-scale social networks implementing DRM-protected ephemeral messages have made off-the-record discussion the norm, and the willingness of courts to force companies to retain device- or server-based logs on a global basis is more and more frequently driving organizations to curate less data and help less. These platforms have appeared in part as counters to corporate and state-backed siloed troves of behavior and demographic data that have at times been weaponized and breached in severe ways.</p><p>Blockchain-based solutions have entered to provide another alternative: transparent data storage and decentralized control. Yet over the last few years, numerous vendors and customers of blockchain technology have reported needing more countermeasures, and the prevailing view from the major central banks and most commercial institutions remains one of rigorously documented blockchain, not bitcoin.</p><p>Several headline-making private blockchains are in development that, rather than focusing on a single aggregated identity for each party as most private chains have done from the start, use some mix of on-disk ring signatures and restricted access to offer still finer-grained access control or perhaps distributed audit of larger databases.</p><p>This work is a contribution to this quadrilateral discussion. There is now a nontrivial low-to-no market for transparent, supportable privacy-preserving technologies, and the intended use cases suggest clear needs for increasing levels of sub-net change flexibility. It is critically important, however, to keep the baby during decisions about the project search space. In real life, making sure that rules are always followed would require all of the implementation staff, which is impossible and might even be illegal. In the same way, the fact that a question can be both specific and general means that it often needs to reveal records and reports that are normally kept secret.</p><p>Also, a strict audit might not find wrongdoing that is hidden with a semi-automatic plausible deniability capacity, but it will stop people from acting unethically when following the rules seems uncertain, undesirable, or not relevant. Balancing the user need for on-chain privacy with the business need for off-chain compliance is a persistent tension that is not insurmountable. Although the clean-cut-and-dried calculus sketched just above appears pro-privacy, the spectrum of intermediate positions is not strictly limited by the extent of technologically feasible deployment. The primary motivator for this broader and less polarizing study, rather, is the way we have come to blankly recite codependent beliefs in a growing multitude of documents. Traditional privacy trust assumptions are collapsing.</p><h3>2. Privacy-Enhancing Technologies</h3><p>The inclusion of the whitepaper demonstrates the growing interest in finding ways to securely enforce compliance on decentralized exchanges. Interesting proposals to enhance privacy in blockchain and cryptocurrencies are flowing out of this activity. There are governance problems in mixing user credentials with keys used in compliance or enforcement schemes, and therefore possibilities for private reneging of the disclosure. In the last few years, efforts and priorities in the development of blockchain technologies and cryptocurrencies have shifted from the creation of “<strong>killer DApps</strong>” to a focus on the enhancement of user privacy. Also, regulators are now more strict about making sure that new crypto applications and services follow rules against money laundering, funding terrorism, and sanctions. This is mainly what drives the creation of privacy-enhancing technologies for centralized crypto infrastructures that are already in place.</p><p>This section explores a number of proposed low-level privacy-enabling mechanisms that, while not a complete list and albeit partial in nature, can be invoked to offer some remedies for privacy on blockchain. They should offer protection at a certain level and spark potential further thought on how to protect users’ privacy even outside the context of decentralized privacy solutions. Specifically: (i) off-chain or state channels; (ii) side chains; (iii) confidential transactions; (iv) zero-knowledge proofs; (v) zero-knowledge contingent or cryptographic payments; and (vi) commitment schemes. In this book, we cover just a few examples of privacy-protection technologies; this is sufficiently comprehensive for the purpose of offering the fashionable version of private payment solutions whose usage is becoming inevitable given the increasing level of scrutiny of exchange regulatory compliance.</p><h3>2.1. Zero-Knowledge Proofs</h3><p>Zero-knowledge proofs have since come a long way, with ZKPs such as zk-SNARKs and bulletproofs being developed in the more recent past. Today, these make up a central part of a privacy-enhancing technology stack called zero-knowledge cryptography. The basic principle of ZKPs is that they allow one party, the prover, to prove knowledge of a fact to a verifier in such a way that the verifier isn’t convinced of the fact, just that the prover knows the fact. Further, the prover reveals nothing to the verifier except for the proof of knowledge itself. The ‘zero-knowledge’ property implies that an interactive proving process does not reveal anything about the prover’s knowledge or secret to any adversary.</p><p>The implication of this property in the context of blockchain is also fundamental: it means that transparency in terms of revealing transaction data can coexist with user privacy. ZKPs find applications in various decentralized systems well outside the scope of privacy coins. For example, they can be used for blockchain systems with unified ledger systems that allow for secure and private voting, defended against forgery, double spending, and vote penalties. <br>Certainly, it is important to note challenges related to the implementation of ZKPs, such as high computational costs and large proofs, as well as the potential for a negative user experience, especially when compared to transparent transactions. Not least, ZKPs could pose a hurdle in terms of on-chain compliance and traceability of transactions, which often run contrary to privacy goals. This does not have to be necessarily the case, though, as applying ZKPs for privacy increases the ability for users to decide how they want to use their privacy. The end goal and ultimate challenge within the space are therefore to ensure that end users are able to have privacy-centric control over their data and do what they want with it.</p><p>Importantly, zero-knowledge technologies offer a very powerful yet fundamental concept to a range of tools and applications that fundamentally value privacy. As such, further development is required for these applications to become more broadly useful. In practice, we need integration with ad-hoc programming paradigms and to understand how they would fit and interact with other technological components. At their most simple technical definition, ZKPs allow a prover to convince a verifier of a mathematical statement without the need to reveal the secret data itself.</p><h3>2.2. Homomorphic Encryption</h3><p>Homomorphic encryption (HE) enables data privacy by encrypting data that can be used to perform computations on the encrypted data. Hence, the data can be manipulated without compromising its privacy. Such an ability is especially important in many applications, such as cloud computing, healthcare, and privacy-preserving machine learning. There are several types of homomorphic encryption, namely partially and fully homomorphic encryptions.</p><p>Partially homomorphic encryption allows for only a specific type of operation, e.g., addition and multiplication, to be done on the encrypted data. Hence, such an encryption scheme can be used for applications that require either of these operations exclusively, e.g., secure integer voting or blockchain transactions. Fully homomorphic encryption (FHE), on the other hand, allows for a sequence of arbitrary operations to be performed on the encrypted data. This turns out to be an extremely powerful tool, as we can construct a “universal” program—a computation that can perform any operation—from that. As a result, we can build very complex algorithms that process inputs without decrypting them at any stage completely.</p><p>Despite such potential, the adoption of homomorphic encryption for complex applications is facing several challenges. First, homomorphic encryption is computationally intensive, meaning that the computing engine must do more operations compared to processing the plaintext equivalent of the operation. Second, not all functions and programs can be translated into a homomorphic computation efficiently. Thus, homomorphic encryption research has been mainly focusing on identifying the right cryptographic primitives and protocols, as well as implementing optimizations to enable practical use cases. Finally, there is a need to perform a robust security analysis to build trust in such cryptographic schemes, especially for sectors that are prone to potential regulatory backlash, such as the blockchain sector.</p><p>Assuming such an encryption is practical (in terms of computation and efficiency) and perfect (in terms of computational security aspects), it can greatly benefit legitimate businesses and authorities that can use blockchain, as it would allow them to comply with data regulations and still perform some computations on the encrypted data. Moreover, integrating homomorphic encryption with blockchain demonstrates the possibility of a hybrid approach for regulators that use programmable privacy systems for the blockchain perimeter.</p><h3>3. Smart Contracts and Compliance</h3><p>Smart contracts are self-executing agreements with the terms directly written into code and stored on a blockchain. Smart contracts automate these agreements entirely, greatly reducing the effort required in administering agreements. Compliance mechanisms may be a part of a smart contract and will likely become common as blockchain-based peer-to-peer systems and platforms continue to grow. <br>As participants interact in these value networks, especially in complicated jurisdictions, the likelihood of interactions needing to comply with numerous regulations from anywhere in the world will continue to rise. Incorporating these compliance mechanisms into smart contracts allows the participants to automate the administration of these potentially expensive additional mechanisms. <br>A critical definition is that these mechanisms must operate on-chain and must operate automatically in accordance with the regulation. Interactions between a user and an application programming interface connected to the blockchain are an important potential threat to external compliance.</p><p>Developing smart contracts that are compliant with various regulations is difficult: the code must function as intended while being interchangeable with other code based on the same regulation, with an identical interpreted outcome. This is even more complex when the regulations concern subjective matters, like who is qualified to register a company, and difficult when regulations are numerous and change often.</p><p>It is well understood that current legal systems are struggling to apply regulations to smart contracts, and the concept has been considered in literature as well as by regulators. While there doesn’t appear to be consensus, there seems to exist an opportunity for proactive development of smart contracts using innovative design and coding practices to create dynamic, adaptable, and compliant systems. Indeed, there are already a number of smart contracts targeted at these on-chain governance and dispute resolution. Specific examples of real-world compliance are: escrow for fiat money sales, regulatory framework according to blockchain-based identity, and regulation of transactions.</p><h3>3.1. Smart Contracts Overview</h3><p>In the simplest terms, a smart contract is a set of rules. It sets forth certain conditions and the outcomes that will follow when these conditions are met and subsequently executed. Smart contracts consist of three basic components: conditions, execution, and outcomes. Given a specific set of conditions, a smart contract will specify the outcomes that one of the parties will implement in a particular way after runtime. The appeal of such a system is that a contract is trustless. The Ethereum Virtual Machine carries out the requirements as well as the transaction&#39;s actual execution.</p><p>According to conventional definitions, a smart contract is executable code that one of the parties will use in accordance with predetermined criteria. It is merely a way of representing contracts between two or more parties through blockchain. These digital and automated contracts are being utilized in a variety of industries to streamline processing and facilitate such tasks as patient data records, financial transactions, last wills and testaments, orders in dispute, and supply chain management. In addition to trust and efficiency, one of the most attractive features of blockchain is the ability to execute smart contracts. Thus, this is expected to facilitate a transformation in the economy with respect to the conduct of automated business processes. However, smart contracts are not without their fair share of challenges when it comes to scalability and security.</p><p>Due to the fact that smart contracts can be difficult to understand because of the lack of user interfaces and actions that operate in the background, the design can be very complex. Additionally, coding errors and program flaws can cause smart contracts to remain vulnerable. Another significant challenge is the fact that there is much uncertainty regarding the legal status and standardization of smart contracts. As such, transactions carried out using smart contracts may not be explicitly illegal at the moment. This means that the law is actively evolving with respect to smart contracts and agreements drawn up through such instruments.</p><h3>3.2. Regulatory Compliance Challenges</h3><p>The topic of regulatory compliance is critically important for smart contract capabilities because, in principle, the ability to encode legal contracts such as option contracts, futures contracts, or more exotic financial instruments into smart contracts is an important value proposition of programmable money. However, building a smart contract platform that can accommodate cryptographic evidence gathering across a number of jurisdictions will be challenging. Moreover, automated contracts are difficult to prosecute, both because it can be difficult to define automated contracts in legal terms and because their execution can lead to unforeseeable results. For example, executed contracts might generate state tax obligations, in which case an executed smart contract may be required to have tax withholding capabilities. It is difficult to understand the full scope of these challenges in advance, in part because smart contracts are a new frontier and regulators’ views on them are still evolving.</p><p>In addition to regulators, developers who build and deploy smart contract technology will also need to worry about regulators. Under existing securities law, firms that aid and abet an unregistered exchange can be held liable, potentially straining smart contract operators who knowingly release software that includes capabilities for violating securities laws. Companies whose agents invest company funds in a non-compliant manner may also be found liable.</p><p>Some of these issues will need to be worked out through case law and perhaps regulation, but some may be addressed at the protocol level. For instance, compliant smart contract platforms might require information for any account that wants to issue a token according to different rules. Note that these various challenges will make it increasingly difficult to use smart contracts to enforce all terms of a contract. One possible response could be hybrid systems that combine smart contracts with human monitoring and oversight that could renegotiate a deal if there were unforeseen consequences from an automated contract.</p><p>An alternative would be for developers to ensure that their smart contracts incorporate all relevant jurisdictions and regulations by default, requiring that any changes to their behavior would need relevant approvals from the participants in their networks. Clearly, there will be a spectrum of risk tolerance for these kinds of transactions, and subject to regulatory changes and legal testing, these issues will require judges or juries to distribute legal liability. A further risk for smart contract adopters is that once different smart contract systems become popular, they might standardize around common practices, leading to de facto forms of centralization.</p><h3>4. Case Studies</h3><p>This section illustrates how Decentralized Identifiers (DIDs) can be used as the primary mechanism to link the identity management and data management layers of Palm with a privacy-preserving subscription protocol.Use cases illustrating how regulatory rules are encoded on-chain with the use of DID verifiable credentials. To further illustrate how an entity can provide additional means to manage the identity and data management on the side, which is useful for organizations that are not necessarily managing their organizational details directly on-chain.</p><p>Assume that Alice wants to subscribe to stream a specific piece of circumstantial data. In Palm, streams are managed through privacy-preserving data handling. This requires the use of pseudonyms, anonymization, pseudonymization, and data obfuscation. To follow the terms of a subscription, Palm must be able to prevent multiple independent data receivers from simply exchanging data between themselves (or even selling access permission) without the original data owners finding out. As a result, users and their devices are identified on stream through the internal generation of volumetric metrics. Specifically, subscribers are able to access the data resources based on a privacy-preserving subscription protocol.</p><p>One example of such a protocol is to issue blinded signatures for each receiving party and to rely on the data owner to trace the entitled data receivers back so that they can sign the data in such a way that access authorization for each specific user is tempting and exclusive. In this example, the data owner can issue a data access key wrapped in the corresponding identity-based decryption private key for the specific attributes of the data receivers so that only these users are able to access the data tokens.</p><p>To work, this protocol requires storage devices performing additional checks when challenges are made. In Palm’s case, data transmissions rely on a reputation system to handle the data provisioning over a decentralized computation substrate. With the blockchain using transaction timestamps as consensusable timestamps, it is challenging for storage devices to challenge the reputations of the actors involved based solely on data transmissions.</p><p>Since storage devices do not gain access to the actual data transmitted from the subscription repository, they have to rely on the reputation of reputation records themselves. Such records are not only performance-based but are also time-bounded so that they can serve as a privacy-preserving alternative for arbitrary access control management. This emphasizes user-defined black-box structures, which can be quite demanding in terms of on-chain contract code execution costs.</p><h3>4.1. Privacy-Preserving DeFi Platforms</h3><p>In many applications, such as DeFi, privacy is a key requirement for users. While many mechanisms for privacy-preserving blockchains have been proposed, they have not been integrated into current popular DeFi platforms because these platforms take a different approach based on public broadcasting of state information as a means of detecting malicious behavior due to human greed. Private smart contracts, i.e., smart contracts that do not store balance information in the contract’s storage, effectively lock an asset indefinitely, preventing the smart contract from detecting malicious behavior or from invoking new privacy-preserving mechanisms at later times. However, contracts can invoke privacy-preserving mechanisms at a later time as long as funds are available. We believe that, in the spirit of empowering individuals, contracts should be able to provide privacy-preserving DeFi mechanisms for individual users. Participants should have the power to convey their funds through this DeFi mechanism and, if desired, to resume the use of any privacy-preserving DeFi functionality.</p><p>If you change existing smart contracts directly, you would not have to use expensive extra instructions or trust-based systems to make sure that deposits, which are important for most DeFi applications, do not include extra information about the funds they hold.</p><h3>4.2. Regulatory Compliance in Supply Chain Management</h3><p>One solution to enforce compliance with local regulations, but also generally to make public some information in a permissioned manner, is to have a government-connected smart contract that can decode permissioned information and check if the upload is compliant or not. This is a rather straightforward solution, but suboptimal in terms of privacy. When parties are allowed to see which data was downloaded, there is the risk of costly re-identification attacks. Known attacks for unsecured IoT can also compromise privacy: crooked IoT gateways can have access to all the blockchain transactions from a considerable number of users and collect a lot of metadata.</p><p>A good solution should also protect the origin and the integrity of the data. Hashes are encrypted client-side and then disclosed. IoT devices download the information and should not be able to change it; they encrypt an authenticated hash of the requested information, and then the smart contract verifies it. This approach is much more complicated as more cryptographic mechanisms are required. However, this solution is secure against both malicious third parties and curious IoT devices.</p><h3>5. Future Directions</h3><p>Developing novel privacy-enhancing technologies. As our digital world evolves and as recent events illustrate the fragility of personal privacy when established systems and institutions are placed under stress, the importance of continuing to develop innovative solutions to address emerging privacy requirements is vital. Concern for privacy is only growing, and so too will the regulatory push for enhanced privacy protection. The application of artificial intelligence and advanced cryptographic solutions raises the possibility of developing more comprehensive and flexible solutions for the protection of individual privacy.</p><p>Regulatory considerations. There is a growing sentiment that privacy and personal data protection standards are being successfully ratcheted up around the world. The bullwhip effect of the regulation has forced multinationals’ hands and generated attempts to directly import aspects of privacy laws into other regions without necessarily undertaking policy development designed to assess their effectiveness. These regulations and associated interest in individual privacy settings are not undertaken in a regulatory-agnostic vacuum. It is a reasonable expectation that regulations will begin to emerge that facilitate on-chain commercial solutions that allow the validation of a participant and the traceability of participants’ actions on a global public blockchain network.</p><p>The integration of privacy solutions on top of public chains. Going beyond novel privacy-enhancing technologies, we expect future work to focus on the integration of these technologies in the wider blockchain ecosystem. Most likely, the majority of further research in the near future will focus on the question of scalability and faster throughput, as well as exploring potential applications beyond cryptocurrencies and considering the ways in which programmable privacy solutions could be integrated into existing blockchain applications.</p><p>Although widespread deployment of privacy-preserving solutions on public blockchains does not meaningfully address all points of regulatory concern, it at least demonstrates that the technological innovation capability to address compliance concerns is almost there. The barriers that remain to be overcome, however, involve all the hard problems around more subtle compliance issues and effective on-chain compliance. As the privacy literature repeatedly points out, meaningful privacy requires ensuring the privacy of every party up and down the stack, including addressing challenging real-world legal issues.</p><p>If user experience and security can’t be scaled with a growing user base, a blockchain will effectively return to being a niche technology. Scalability concerns involve a host of other significant challenges, agility with asset creation, security, private computation, cross-asset, cross-system transactions, etc., the resolution for which will also prove to be challenges in the short to medium term.</p><h3>5.1. Emerging Technologies</h3><p>Advancements in technology continue to improve privacy preservation within blockchain transactions. Integrating the technology of decentralized identity with current public ledgers will altogether fashion programmable privacy within transactions. The nature of decentralized identity in permissionless public ledgers can redefine the privacy levels of existing schemes. For instance, if a loan contract of a banking business can access the credit score of a loan applicant only, it can prevent further infringement of personal data. At the same time, emerging technologies indeed play a notable role in this domain. For example, privacy-preserving computation can maintain the confidentiality of data and the secret key by encrypting sensitive information. In terms of compliance, advances in machine learning and artificial intelligence can allow the mining of relevant user details to allow for evidence gathering and utilization thereof.</p><p>Despite these advancements, there are some issues regarding technical immaturity, lack of standards and guidelines, costly technology, and nonlinearities. It is noted that there is no guarantee that these emerging technologies will sustain and evolve given the difficulties of being part of the current legal regulations and policies. Although modern privacy technologies might be accepted as new additions, their degree of acceptance by the blockchain industry is a real concern.</p><p>However, it is suggested that risks should be managed proactively by considering possible legal actions, social arrangements, and technology standardization and regulation. Redefining the SPIRAL model, the privacy domain continues to advance. Now, in addition to the initial three stages, Privacy Desire and Current Control, the definition of the SPIRAL Model has embraced the fourth stage: desired state — politico-regulatory and compliance-type innovation. It is the first instance of a priori theoretical politics fully driving privacy research development.</p><h3>5.2. Regulatory Trends</h3><p>In February 2022, the final rule on records and reports concerning transactions in digital assets came into force. It extended the Anti-Money Laundering (AML) requirements on unhosted wallets to intermediaries between cryptocurrency transactions. This policy regulating the transmission of virtual currencies effectively discriminates against and could even “kill” privacy in digital assets. In many countries, the amendments have refueled heated debates about — and even highlighted how — a patchy privacy tradition undergirds and exacerbates society’s lack of generalized support for fungible pseudonyms. Businesses must navigate the complex and often conflicting regulatory environments in multiple jurisdictions. Some jurisdictions have a reasonably favorable stance towards the use of privacy-enhancing technologies.</p><p>The recent policy excerpt reads: “Digital assets. Privacy-focused and privacy-specific solutions such as zero-knowledge, selective disclosure, and adding metadata could be part of the solutions for new solutions for positing identity.”</p><p>In response to the idea of privacy essentiality, regulators are increasingly moving away from a binary approach to privacy and toward a gradient approach. Regulated entities are starting to lean in more proactively on their lobbying efforts to shape regulations favorably, with a corporate privacy group being a case in point. The development supports the expectation that an increasing number of projects will be launched favoring solutions that possess a “privacy switch,” capable of flipping between regular mode and privacy mode. Project managers and protocol teams should make themselves aware of relevant global and local regulatory changes affecting privacy and chainspace dispute resolution.</p><p>It is still too early to draw definitive conclusions about the long-term privacy ecosystem implications of the transmission of virtual currency amendments. Similar precedents, such as the e-Privacy Regulation, are yet to become operational. A lot of the ways that blockchain and privacy policies interact right now are limited to centralized verification of personal data. But the new policy has effects on businesses and customers that use the protocol, because failing to verify off-chain information comes with penalties like liquidity management fees and extra losses from &quot;premium&quot; fees changing the demand for conversion.</p><p>Increased compliance burden is also known to depend on the private key model. The rebound of privacy-focused coins following the implementation of travel rules that require to collect and share pertinent transaction information with counterparties is also, as yet, a little explored area of research. How viable would proposed privacy-protective solutions be tested in chain to reduce the cost of transaction reporting? In the past, chains overdosed. Suggestions include attracting funding from non-governmental organizations to deal with higher registration expenses as an alternative to costly criminalization. Rebate-based business models improve monetary inclusiveness. A small study explores the sources and types of anonymization available to address specific compliance requirements in greater detail than likely possible for this blockchain and privacy topic. The emergence of compliance-focused KYC also initially increased the relative demand for technologies and solutions. A greater number of comprehensive demonstrations of smart contract privacy and accessibility, or indeed, the interactions with the physical world, would provide a fuller understanding of the arguments.</p><h3><strong>Final thoughts</strong></h3><p>The intersection of privacy and compliance is increasingly important in today’s digital economy, particularly as blockchain technology evolves. Striking a balance between privacy and compliance is a complex but necessary challenge, and advances in privacy-enhancing technologies offer a promising path forward. Tools such as zero-knowledge proofs, homomorphic encryption, and privacy-enabled smart contracts offer solutions that address these dual concerns, but they also introduce new hurdles, particularly in terms of scalability, user experience, and regulatory acceptance.</p><p>As we move towards a future where privacy is programmable and compliance is on-chain, continued collaboration between regulators, technologists, and businesses will be essential. These efforts will help create a regulatory framework that both protects individual privacy and ensures legal compliance, fostering innovation without sacrificing trust. Ultimately, the development of privacy-enhancing technologies will not only protect personal and corporate data, but also lay the foundation for more secure and compliant decentralised systems.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=c4576c681dec" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[An In-Depth Analysis of the Knapsack Problem]]></title>
            <link>https://medium.com/@0x_Rorschach/an-in-depth-analysis-of-the-knapsack-problem-1d13714e3ead?source=rss-0b668e3fef6c------2</link>
            <guid isPermaLink="false">https://medium.com/p/1d13714e3ead</guid>
            <category><![CDATA[problem-solving]]></category>
            <category><![CDATA[algorithms]]></category>
            <category><![CDATA[innovation-games]]></category>
            <category><![CDATA[knapsack-problem]]></category>
            <dc:creator><![CDATA[Rorschach]]></dc:creator>
            <pubDate>Tue, 10 Sep 2024 21:07:56 GMT</pubDate>
            <atom:updated>2024-09-10T21:07:56.859Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*qVZUkcyGY5pathYr93AN3Q.png" /></figure><h3>1. Introduction to the Knapsack Problem</h3><p>The Knapsack Problem is a classic optimization problem that seeks to understand the best way to fill a finite space to maximize a certain outcome. Typically, the problem is conceptualized through resource allocation and involves an individual trying to maximize the value of items they possess without exceeding an available resource. This kind of problem is useful in a variety of scenarios where limited space or budgets are common constraints. Formally defined, a set of items exists, each with a corresponding weight and value. Additionally, the knapsack has a defined capacity that cannot be surpassed. The goal is to maximize the total combined value for all items chosen without surpassing the weight limit. (1)</p><p>A variety of fields have historically contributed to work on the knapsack problem, including mathematics, business, economics, and computer science. At its heart, the knapsack problem represents a combinatorial optimization challenge, meaning that a solution needs to be selected from a finite number of combinations that exhibit a certain collection of characteristics. Combinatorial optimization helps determine the best course of action for an individual or entity when presented with options. Simple examples illustrate the concept of the problem and its solution using basic definitions and prize ‘loot.’ The practical relevance of the problem is pervasive; many real-world constraints can be translated into a ‘knapsack’ or ‘limited space’ challenge. These constraints can be qualitative or quantitative. (2)</p><h3>2. Types of Knapsack Problems</h3><p>Knapsack problems come in different types based on constraints and the nature of decision variables. These types include the 0/1 knapsack problem, fractional knapsack problem, simple knapsack problem, bounded or unbounded knapsack problem, multiple knapsack problem, and multi-dimensional or multi-choice or binary knapsack problem. In this memory-effective but more or less comprehensive overview of the knapsack problem, the focus will remain on the 0/1 knapsack problem and fractional knapsack problem. (3)</p><p>The 0/1 knapsack problem constitutes the issue of what combination of items should be included in the bag to maximize the value within the given capacity, in which an item is either fully included in the bag or does not appear in the bag. As such, the 0/1 knapsack problem is an instance of the integer programming problem. On the other hand, the fractional knapsack problem, also known as the simple knapsack problem, is the issue of what parts of items one should place in the bag to obtain the maximum value, in sharp contrast to the 0/1 knapsack problem, in which one could only choose none or all of a given item. Because of the unique structure or pattern of this problem, one could obtain a very efficient algorithm for solving it. Also, the minimum problem regarding how to make use of the capacity is still the fractional knapsack problem. The same as the 0/1 knapsack problem, the fractional knapsack problem can be easily formulated via a linear programming model. (4)</p><p>It is widely believed that fractional knapsack problems are easier to solve than 0/1 knapsack problems, especially in terms of minimizing the complications of the phase of solving the problems. However, the fractional knapsack problem is not suitable for real-world situations in all cases. For example, when an item is characterized by considerable indivisibility, the situation of the ordinary knapsack problem is more or less represented. As a result, merely easier solving of fractional knapsack problems does not always mean that fractional knapsack problems can be replaced by ordinary knapsack problems. This attribute leads to the consideration that the fractional knapsack problem uniquely reflects the essentials of practical situations only to a considerable extent. Needless to say, the fractional knapsack problem is important in its own right. (5)</p><h3>2.1. 0/1 Knapsack Problem</h3><p>The 0/1 Knapsack Problem can be defined as follows: given a set of items, each with a weight and a value, determine which items should be selected in order to maximize the total value without exceeding the capacity of a knapsack. In the 0/1 Knapsack Problem variant, each item can either be included in (0) or excluded from (1) the knapsack. This discrete choice makes the problem an NP-hard optimization problem. A variety of techniques have been applied to solve it, ranging from integer programming to genetic algorithms. The most well-known approaches make use of slow but exhaustive power set enumeration, and it is a core feature of greedy algorithms. Nevertheless, the 0/1 Knapsack Problem is most commonly considered in the context of dynamic programming (DP), a powerful approach that has resulted in algorithms with computational complexities that range from O(n) to O(nW).</p><p>Below are examples of instances of the 0/1 Knapsack Problem. The 0/1 Knapsack Problem has several practical applications and has been extensively studied. Classic dynamic programming algorithms have been proposed, and DP algorithms can be employed at larger scales with the help of memorization (the top-down approach). Moreover, there exist a number of heuristics that exploit certain properties of the optimal solution, and they can be used to speed up standard DP algorithms. (5)</p><h3>2.2. Fractional Knapsack Problem</h3><p>Although the Fractional Knapsack Problem may appear as a less strict version of the original 0/1 knapsack problem, it can more appropriately be seen as a problem in its own right when making the full development. The main difference with the associated problem is that for the fractional version, we are allowed to take an item partially. In other words, we can select any fraction of the object instead of its totality only. This flexibility at the level of the objects often helps to develop simpler and faster algorithms. There is indeed a greedy strategy that solves the fractional knapsack problem by considering solely the value of the items instead of having to account for their weights. The usual strategy consists in considering the value-to-weight ratio, which we know how to maximize. The remainder of this document can be spent on detailing how the actual fractions are to be selected. For the specific case when the items cannot be divided, we are not obliged to restrict ourselves to greedy strategies, and a dynamic programming solution can hence be provided. (6)</p><p>More technically, provided that as many fractions as required are at our disposal, each fraction of an object can independently be determined. Indeed, it stands to reason that the fraction of the largest value-to-weight ratio should be taken until either the weight of the remaining fraction becomes more than necessary or when the available space in the knapsack has its limit reached. As for the applications, the fractional knapsack problem can, among others, describe appropriate stock management, optimal ways of loading non-divisible items into a vehicle, or the selection of the best investment when partial selections are possible. From an algorithmic point of view, if the best algorithmic complexity can in general be expected to be linear with respect to n, the binary character of the knapsack problem often allows for a simpler solution. (7)</p><h3>3. Approaches to Solving the Knapsack Problem</h3><p>Various approaches have been suggested to find solutions to the Knapsack Problem. The effectiveness of these diverse methods is generally the function of the constraints they try to overcome. For example, dynamic programming represents a powerful method for solving the 0/1 Knapsack problem. This is due to the recursive nature of the solution that makes it suitable for this type of knapsack only. In contrast, greedy algorithms are more suitable for solving the fractional knapsack problem, as their solution usually works in a situation where fractional weights are permissible. Greedy algorithms sort all objects based on value and choose content based on their total weight. The branch and bound method is yet another method that provides an optimized solution to the Knapsack Problem. It initially generates the possible solution by uninformed searching, rapidly finding the infeasible solution and pruning impossible solutions. (8)</p><p>Each method has its own emphasis on certain factors. Greedy algorithms and dynamic programming give precedence to objects with high total value assigned. Dynamic programming chooses objects using capacity filled within the sack, while greedy algorithms consider objects using different weight increments. While dynamic programming does not provide the best fractional solution, greedy algorithms provide the best fractional solution without any problem. All of the above-listed solutions have been applied to a numerical example. In practice, it will be difficult to choose which method to use for solving the Knapsack Problem based on the problem’s situation. The possible methods are presented by applying three different conditions for the same given input. However, this result is analyzed and thus highlighted. 9</p><h3>3.1. Dynamic Programming</h3><p>In the following sub-section, we will understand dynamic programming, which proves to be an optimal approach to solving the 0/1 Knapsack Problem and also the basis of our detailed study of the same in upcoming sub-sections. Dynamic programming is a mathematical foundation that dissects problems by breaking them down into simpler, overlapping sub-problems.</p><p>Step 1: First, we set the initial values as zero for both the matrix defined.</p><p>Step 2: We fill the matrix while referring to the previous entry values in the same matrix.</p><p>Step 3: Formulate the final solution by referring to the last filled entry in the matrix.</p><p>Pros and Cons: (i) It follows a systematic approach, and the utility of the way ensures that this is the optimal solution. (ii) The biggest drawback is that it requires a higher amount of computational resources, as we will see that the space and time complexity of the dynamic programming approach is higher.</p><p>Drawbacks: (i) We need to pre-plan the storage requirement since, being an array substitution, we can’t expand the storage unit when needed. (ii) The execution times of for-loops are also fixed and the same for all dynamic programming manipulations, which are computed using the formula, resulting in a fixed output of the type. 10</p><h3>3.2. Greedy Algorithms</h3><p>Besides the optimal solution, the Knapsack Problem can be solved via alternative solution approaches as well, such as greedy algorithms. They are a collection of algorithms designed to solve optimization problems by following the best available immediate payoff decision. The objective of these algorithms is to choose the alternative that is a local optimum. Additionally, the process analysis of greedy algorithms reveals that, at every step, the decision taken represents the best one from the current viewpoint. This particular characteristic is what separates greedy algorithms from other classes of optimization algorithms. (11)</p><p>These properties of greedy algorithms make them particularly appealing for some problems where we make selections between fractional types; this is a class of non-boolean selections. The main idea behind greedy algorithms is to select the amount of each selected item in such a way that it can be most advantageous. After considering them step by step, we can see that these types of algorithms are pretty coarse, straightforward, and efficient in computation for solving non-integer problem areas of combinatorial problem alternatives such as resource allocation problems. The general steps of selecting choices are serious and applicable at the beginning level. The algorithms are capable of aggregating greedy-negative selection level contraries and using them in the mathematical programming model or alternatives of complex decision systems. Constituted small parts in the area of combinatorial optimization have not been adequately researched in the literature. These weak points prevent the creation and application of the relationship between the selection procedure. (12)</p><h3>3.3. Branch and Bound</h3><p>Branch and Bound For certain NP-hard 0–1 Integer Programming Problems like the 0–1 knapsack and the Set Covering Problem, a combination of the improved heuristics suggestions with a mathematical derivation through the set partitioning could lead to good results. The first thing we consider for a new heuristic is a K-setting because it is a group with a fixed number of activities relevant to the customers. The K-set heuristic is designed to receive a table with the set of K items from a new class and their quantitative parameters. We then enter the K-set into our previous cover category, which was also created from the attribute-sort previously unclassified objects. We must now ensure the correct location of the new items and create K new K-sets.</p><p>There is no assurance of the global optimality of the GAP and the Set Partitioning Algorithms. For this problem, there is a rather dominant method that offers the optimal solution compared to other exact algorithms, which are not able to solve the problem with complex relations. The Branch and Bound has the ability to obtain the best solutions in shorter time than other accurate algorithms.</p><p>This algorithm was developed to assist in the exact solution of the 0–1 integer programming problems. The method applies to small problems. This algorithm defines branches at each solution with restrictions that are relaxing while it fixes one or more variables. If the solution with the remaining restrictions is requested and proven to satisfy a minimization problem, then it becomes a frontier solution; otherwise, the solutions are deemed infeasible. (13)</p><h3>4. Real-World Applications of the Knapsack Problem</h3><p>The report yielded a comprehensive exploration and assessment of the Knapsack Problem. In addition to outlining its definition and purpose, it effectively examined the different types of the Knapsack Problem and details of its complexity. Case studies and comparative analyses aided in illustrating the different approaches, algorithms, and real-world applications of the Knapsack Problem. Further insights revealed instances of automated refinements and the most recent research activities within this framework. However, the report deems that real-world cases have some limitations in their investigation of the problem, suggesting that future research could address this shortfall. Furthermore, the analysis of the existing literature implies that the supply chain is a very promising new area for studying the unbounded Knapsack Problem of clustered items, as seen in the case studies. (6)</p><p>The Knapsack Problem, in all of its various forms, has important real-world applications. In logistics and resource management, the decision of how to select items involves optimization, costing in both money and time. In finance and investment, potential revenue or earnings from investments need to be maximized. In staffing problems, employees become the items with their qualifications and salaries as weight and cost or benefit. The major project to which they can be assigned will be the constraint. Either the cost of assigning them to a major project or the skills required for the project and possessed by the employees are aspects considered. These types are defined with respect to what the constraint is and what exactly is being maximized in accordance with that restraint. In the Knapsack Optimization Problem, the main focus is to identify how the problem extension is connected to existing relatable problems. (14)</p><h3>5. Conclusion and Future Research Directions</h3><p>In this work, I analyzed the Knapsack Problem (KP), an optimization problem that addresses various real-world scenarios. The focus was on the classic 0/1 KP and its solution using dynamic programming, greedy algorithms, and branch and bound approaches. I also explored the fractional KP and 0/1 KP with profit only, as a special case of the knapsack problem. The KP falls under the NP-Hard category, with exact algorithms presenting solution complexity of O(2^n) for the worst case, while heuristics offer near-optimal solutions using a non-exhaustive method in O(n), where n is the number of instances. These characteristics are significant, as the KP cannot be classified into a specific problem type solely based on the number of constraints and decision variables. The key takeaway is that each problem type requires a solution methodology that best fits the instance characteristics.</p><p>As a further development of this study, I propose several research opportunities. New optimization techniques can be applied using deep learning models that leverage features of the decision variables, improving computational time. The integration of various methodologies into a hybrid approach can combine the strengths of each strategy. The practical nature of the KP is vital, and its effectiveness can be explored across multiple industries. Therefore, investigating the presence and application of the KP model in sectors such as energy, healthcare, and services could lead to innovative studies and a new way of addressing the existing problem.</p><p>Lastly, the increasing complexity of the KP makes it an appealing topic for optimization studies. Larger and more intricate instances typically demand the use of more explicit methods. Given the current challenges and advances in optimization, developing exact and faster solutions could generate new alternatives in addressing real-world issues.<br> <br> <br><strong> References:</strong></p><p>1. Pferschy U, Nicosia G, Pacifici A, Schauer J. On the Stackelberg knapsack game. European Journal of Operational Research. 2021 May 16;291(1):18–31. <a href="https://www.sciencedirect.com/science/article/pii/S0377221720307931">[HTML]</a></p><p>2. Zeynali A, Sun B, Hajiesmaili M, Wierman A. Data-driven competitive algorithms for online knapsack and set cover. InProceedings of the AAAI Conference on Artificial Intelligence 2021 May 18 (Vol. 35, №12, pp. 10833–10841). <a href="https://ojs.aaai.org/index.php/AAAI/article/view/17294/17101">aaai.org</a></p><p>3. Patel D, Khan A, Louis A. Group fairness for knapsack problems. arXiv preprint arXiv:2006.07832. 2020. <a href="https://arxiv.org/pdf/2006.07832">[PDF]</a></p><p>4. Jin C. 0–1 knapsack in nearly quadratic time. InProceedings of the 56th Annual ACM Symposium on Theory of Computing 2024 Jun 10 (pp. 271–282). <a href="https://dl.acm.org/doi/pdf/10.1145/3618260.3649618">acm.org</a></p><p>5. Moradi N, Kayvanfar V, Rafiee M. An efficient population-based simulated annealing algorithm for 0–1 knapsack problem. Engineering with Computers. 2022. <a href="https://link.springer.com/article/10.1007/s00366-020-01240-3">[HTML]</a></p><p>6. Cacchiani V, Iori M, Locatelli A, Martello S. Knapsack problems — An overview of recent advances. Part II: Multiple, multidimensional, and quadratic knapsack problems. Computers &amp; Operations Research. 2022 Jul 1;143:105693. <a href="https://www.sciencedirect.com/science/article/am/pii/S0305054821003889">sciencedirect.com</a></p><p>7. Sun B, Zeynali A, Li T, Hajiesmaili M, Wierman A, Tsang DH. Competitive algorithms for the online multiple knapsack problem with application to electric vehicle charging. Proceedings of the ACM on Measurement and Analysis of Computing Systems. 2020 Nov 30;4(3):1–32. <a href="https://arxiv.org/pdf/2010.00412">[PDF]</a></p><p>8. Anditta M, Amartya N, Warnars LS, Warnars HL, Ramadhan A, Siswanto T, Mantoro T, Noordin N. Dynamic programming algorithm using furniture industry 0/1 knapsack problem. In2023 5th International Conference on Cybernetics and Intelligent System (ICORIS) 2023 Oct 6 (pp. 1–6). IEEE. <a href="https://ieeexplore.ieee.org/abstract/document/10352182/">[HTML]</a></p><p>9. Gupta V. Greedy algorithm for multiway matching with bounded regret. Operations Research. 2024. <a href="https://arxiv.org/pdf/2112.04622">[PDF]</a></p><p>10. Vieira T, Cotterell R, Eisner J. Searching for more efficient dynamic programs. arXiv preprint arXiv:2109.06966. 2021. <a href="https://arxiv.org/pdf/2109.06966">[PDF]</a></p><p>11. Tang J, Tang X, Lim A, Han K, Li C, Yuan J. Revisiting modified greedy algorithm for monotone submodular maximization with a knapsack constraint. Proceedings of the ACM on Measurement and Analysis of Computing Systems. 2021 Feb 22;5(1):1–22. <a href="https://dl.acm.org/doi/pdf/10.1145/3447386">acm.org</a></p><p>12. Zhang T, Banitalebi-Dehkordi A, Zhang Y. Deep reinforcement learning for exact combinatorial optimization: Learning to branch. In2022 26th international conference on pattern recognition (ICPR) 2022 Aug 21 (pp. 3105–3111). IEEE. <a href="https://arxiv.org/pdf/2206.06965">[PDF]</a></p><p>13. Braik M, Ryalat MH, Al-Zoubi H. A novel meta-heuristic algorithm for solving numerical optimization problems: Ali Baba and the forty thieves. Neural Computing and Applications. 2022. <a href="https://www.researchgate.net/profile/Malik-Braik/publication/353781024_A_novel_meta-heuristic_algorithm_for_solving_numerical_optimization_problems_Ali_Baba_and_the_forty_thieves/links/61a640bd743d9629db2718d3/A-novel-meta-heuristic-algorithm-for-solving-numerical-optimization-problems-Ali-Baba-and-the-forty-thieves.pdf">researchgate.net</a></p><p>14. Ozsoydan FB, Gölcük İ. A reinforcement learning based computational intelligence approach for binary optimization problems: The case of the set-union knapsack problem. Engineering Applications of Artificial Intelligence. 2023. <a href="https://www.sciencedirect.com/science/article/pii/S0952197622006789">[HTML]</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1d13714e3ead" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Role of Synthetic Data in Machine Learning]]></title>
            <link>https://medium.com/@0x_Rorschach/the-role-of-synthetic-data-in-machine-learning-ef67a5d9b0fc?source=rss-0b668e3fef6c------2</link>
            <guid isPermaLink="false">https://medium.com/p/ef67a5d9b0fc</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[synthetic]]></category>
            <category><![CDATA[data]]></category>
            <dc:creator><![CDATA[Rorschach]]></dc:creator>
            <pubDate>Sat, 07 Sep 2024 09:53:48 GMT</pubDate>
            <atom:updated>2024-09-07T09:53:48.703Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="Credit goes to gretel.ai" src="https://cdn-images-1.medium.com/max/1024/1*UfA_bWJJsisKtlJAX0gimg.png" /></figure><h3>1. Introduction</h3><p>Given the recent developments in the field of machine learning, particularly deep learning, which depend upon large quality and salient data, synthetic data generation could play an important role in the future of machine learning. This essay is a primer on synthetic data generation in the field of machine learning, especially deep learning. We discuss its various applications and the most recent advancements in the topical issue of using deep learning for synthetic data generation.</p><p>Data scarcity is a major issue in the training of machine learning models, especially deep learning models. In the real world, high-quality data to train such models is scarce, and the ground truth data needed to validate and test these models is even scarcer. Generation of synthetic data is a promising solution to tackle these issues. The most significant advantage of using synthetic data is that we have access to an unlimited amount of data, which could help in training large deep learning models that require a plethora of data to the point of satisfying the capacity of such models. Synthetic data is also useful in training models for performing tasks that have certain privacy concerns or ethical issues, such as crowd analysis, face identification, and healthcare-related tasks.</p><h3>2. Understanding Synthetic Data</h3><p>Super simple. In the simplest of senses, we might just say synthetic data is an object or dataset that is generated using certain rules or criteria. We can also mention immediately that it is usually done with computers to contrast with naturally or ‘actually existing’ data, which may be characterized by existing in the ‘real’ world, its cost of gathering, or the uncertainty coalescing with its generation.</p><p>After the model of the data-generating process (DGP) is agreed on, then it is just a matter of running procedures using the model to have produced data. In one implementation of this process to get synthetic data similar to actual existing data, one could feed directly in the ‘actual existing’ data to the procedure and receive data that would be, within the limits of randomness, almost indistinguishably similar to the actual existing data.</p><p>In sum, synthetic data is entirely dependent on how it was created. It can have no relationship with actual existing data other than the times when it is designed and developed to mirror actual existing data. It can be directly based on actual existing data so that it tries to be similar to the actual existing data. It can be partially based on actual existing data and partially not.</p><h3>3. Generation Techniques</h3><p>Synthetically generated data has been established as a valid replacement or supplement to real-world data in various areas of life, primarily in the context of machine learning. The Generation Techniques section provides an overview of methods used to create synthetic data designed to mimic the properties of real-world data. When combined with real-world data through techniques like data augmentation, synthetic data can reduce the learning algorithm’s reliance on real-world examples and ultimately improve classifier performance. This section shows the necessity of not only using high-quality synthetic data to accurately reflect the properties of the real-world distribution but also generating high-quality data. Poor data quality happens as frequently in purely synthetic data as in purely real-world examples.</p><p>The use of synthetically generated data goes hand in hand with data quality. Guaranteeing the accurate reflection of real-world data properties in the synthetic data is just as important in the generation of synthetic data as in real-world data. Guaranteeing data quality can be seen as the missing interpretation of the black box of synthetically generated data. For machine learning tasks with supervised learning, such as image recognition, synthetic data can be used as an exclusive or supplemental substitute method when there is a shortage of labeled examples predominantly generated by domain experts. However, learning on exclusively synthetic data can result in models only performing well on synthetic data, with realistic behavior suffering since the synthetic examples do not accurately reflect the real-world distribution. Six parameters required for synthetic generation are then explicitly classified: domain-specific, global, sampled, network-specific, real-world-specific, and auxiliary data-fetching ways.</p><h3>3.1. GANs (Generative Adversarial Networks)</h3><p>Ian Goodfellow first introduced the technique known as &quot;Generative Adversarial Networks&quot; (GANs) in 2016. A short description of a GAN is as follows: a generator is drawing fakes out of random noise, and the discriminator learns to distinguish between real and fake images. The key of the GAN is in the adversarial collaboration between the generator and the discriminator. The generator iteratively produces samples that try to maximize the error of the discriminator, making it assign high probability to fakes. By learning to discriminate between real and fake data, the discriminator indirectly learns the distribution of the data. It also acts as a supervisor for the generator, guiding it toward producing more realistic samples.</p><p>Since the introduction of GANs, there have been many adaptations and improvements to the technique, substantially broadening its use. There are many names of derivatives of GANs, and I don’t intend to review these here. The original idea behind the generative adversarial networks was to generate images. Since then, in the years that followed, GANs were also used to generate audio based on pre-recorded audio samples, create imaginary faces that don’t exist, create objects that don’t exist, and many more. But GANs can be used for creating any type of data. This contrasts with the variational auto-encoder (VAEs) framework, which can also be used to generate images or texts, but their main use is to generate new samples with desired features. In the standard GANs, the generator can generate basically everything within the domain of the training data. That flexibility comes with both advantages and disadvantages.</p><p>The impact of GANs led to the evolution of their use from GANs used as a photo generator for generating fake pictures of cats and dogs to a strong focus on the generation of synthetic data for machine learning tasks. The main goal of synthetic data is to generate artificially a large amount of realistic and diverse data points across the input space of the machine learning model that would help us have a better prediction, particularly in adversarial robustness. For now, GANs are considered the most convenient approach for synthetic data generation. Besides GANs, you can train a standard neural network whose output is noise on normal labeled example inputs. Due to overtraining on fake data, the network is then capable of generating new examples.</p><h3>3.2. Variational Autoencoders</h3><p>Although GANs are widely used to produce synthetic data, variational autoencoders (VAEs) are among the most well-known practical methods for this task. The autoencoder architecture is composed of an encoder that maps the input to a fixed-size latent space and a decoder that translates the latent into a reconstruction. The VAE model builds on the classic autoencoder idea by learning a probabilistic latent space. Both the encoder and the decoder can learn the latent variable model that is defined by the way the observations are generated. VAEs consist of a neural network-based encoder and decoder that parameterize a mean and a variance in the latent space. It is applied to make sure the latent representation resembles a data distribution. After the training, we can sample data in the latent space and decode it into reconstructions, thus generating synthetic data. Moreover, the normally distributed prior makes the method generate natural examples from the data distribution.</p><p>Variational autoencoders (VAEs) are often used instead of GANs to define generative models in unsupervised domains. This is especially true in health informatics, where the high-dimensional data has heavy-tailed distributions and is not as well suited to simple noise-masked reconstruction processes, which are what VAEs do. VAEs have been used to generate synthetic ECG waveforms like GANs for data augmentation with promising results. Mean-field variational inference is applied to the problem of inferring the latent values. The goal is to find the closest match (in terms of a distance metric that could be, e.g., the KL-divergence) between the real distribution and the “learned” distribution that is factorized, thus easier to compute. A downside is that this solution minimizes the KL-diversity without considering the ability of VAEs to generate reconstructions that are close to the real data, since these ones are based on the learned generator. Thus, different training objectives have been proposed-such as the adversarial autoencoder-that allow improving the reconstruction property of the artificial samples of VAE.</p><h3>4. Applications of Synthetic Data in Machine Learning</h3><p>Much of the focus of synthetic data in the recent machine learning literature has been on using generative models such as normalizing flows and GANs to produce artificial examples of specific labeled distributions. In this setting, two main applications for synthetic data are identified. Firstly, synthetic data is used as an augmentation to the training set when a small dataset is available. This acts as a regularizer on the hyperparameters and can prevent overfitting in ill-conditioned models. Secondly, synthetic data is used as a privacy-preserving synthetic substitute for human subject data. One solution to protecting the privacy of an individual in government statistics and surveys, medical research data, or consumer credit data is to share the statistics encapsulated in the dataset rather than the data itself. Nonetheless, information leakage in shared statistical rates and structure can make it possible to go back from the released statistics to the individual data samples. Instead of sharing the samples, something “like” the samples, i.e., synthetic resamples of the individual data, is provided as synthetic data to third parties.</p><p>Generative models that replicate the data to a certain extent can then be used to infer on the generative distribution instead of the private one and answer questions about it. For instance, in medical research, third parties can use synthetic data to address causal issues with public skepticism regarding the population distribution, improving the transparency of the resulting findings. Overall, research on evaluated generative models for synthetic resampling suggests that the accuracy of the replication is currently highly dependent on the type of data and its characteristics. Additionally, despite the advances in generative models, the realism of the whole structure of the synthetic data has been mainly illustrated in highly stylized examples with a small number of attributes. This suggests that synthetic data is currently only of potential use rather than direct practical application for most research projects.</p><h3>4.1. Data Augmentation</h3><p>A key component to the ability of modern machine learning models to generalize well to unseen data is a large training dataset. This typically requires costly and time-consuming manual annotation of examples that are representative of the domains in which the model is expected to operate. However, one can effectively expand datasets by employing synthetic data. This practice has found success in applications such as adversarial robustness training, domain adaptation, and few-shot learning and has been shown to be a practical approach to improving sample efficiency. In the context of computer vision, common data augmentation techniques include rotational invariance, horizontal and vertical reflection, and random cropping. Exposure of deep learning models, including convolutional neural networks, to several kinds of synthetic data has been shown to boost performance. Several approaches to employing synthetic data to pre-train generative models have shown great potential in computer science training, often resulting in state-of-the-art performance.</p><h3>4.2. Privacy-Preserving Data Sharing</h3><p>Machine learning algorithms can take advantage of the vast pools of personal and proprietary data in their training sets. Therefore, data sharing is essential in the field. Especially for collaboration among various organizations, much sensitive and valuable information is worth hiding and should not be disclosed. Synthetic data could offer an approach for secure and privacy-preserving data sharing in privacy-preserving machine learning applications. For example, the data reporter could report synthetic data to the public while keeping the raw data secret. By leveraging synthetic data, the numbers in the raw data could be described in a statistically correct manner or even worse, but the actual values in the raw data can be kept private.</p><p>The non-iid data is barely uniform and has too many outliers, resulting in a biased model that cannot be applied to practical applications. Therefore, a sophisticated transformation of synthetic data is expected to correct bias and keep the data private at the same time. This mechanism could alleviate the non-iid issue to make the synthetic data have a uniform distribution and fewer outliers. On the other hand, it can reduce the influence of non-iid data on data privacy, which is promising in real-world data applications. The tremendous success also exposes the non-iid data privacy threats and privacy challenges in big data and machine learning. The privacy-preserving research in the deep learning domain is at the theoretical research stage and has yet to be fully supported in practice.</p><h3>5. Challenges and Limitations</h3><p>For large applications of machine learning (ML) and deep learning (DL), it is difficult to provide enough training data in the right manner due to some privacy and legal issues. However, it is necessary to ensure the reliability and accuracy of ML/DL models. Synthetic data has been proved useful in providing enough training samples to largely improve the accuracy of ML/DL models in some practical applications. A crucial task is to evaluate the performance of ML/DL models on synthetic data. Reviewing the current research on the role of synthetic data in ML, this paper conducts a qualitative evaluation on ten prevailing machine learning tasks.</p><p>Although synthetic data has shown its potential in machine learning (ML) tasks, the extensive application of synthetic data still faces several issues and challenges. The first thing that cannot be ignored is that synthetic data might induce possible bias between the DL model and the real world, as well as break the ethical constraint. In addition, the generation of high-quality synthetic data with fidelity and variability is still technically challenging to some extent. This problem is even aggravated by some real-time machine learning tasks, especially in high-dimensional feature and label spaces, where model-based synthetic data may not ensure the sufficiency of the training set. Level of difficulty for certain synthetic data generation is another limitation, which is that it is rather intricate to define a suitable data sampling strategy for certain ML tasks. In addition, the computational cost is also a constraint in some cases, as generating synthetic data needs to train GANs or other generative model iterations, which are computationally expensive in some sense.</p><h3><strong>6. Conclusion</strong></h3><p>In summary, in this article i examined the role of synthetic data in enhancing machine learning models, particularly in addressing data scarcity and privacy concerns. While synthetic data offers promising solutions, challenges such as data fidelity, bias, and computational costs still persist. Ongoing research and advancements in generative techniques, like GANs and VAEs, will continue to push the boundaries of synthetic data’s application in various fields, including healthcare and security-sensitive domains. As these technologies evolve, their ability to create high-quality, reliable data will be crucial in driving the next phase of machine learning advancements.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ef67a5d9b0fc" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Entropy and Machine Learning]]></title>
            <link>https://medium.com/@0x_Rorschach/i-was-watching-key-equations-behind-probability-and-wanted-to-prepare-an-article-on-entropy-in-a9c9d0bc58b5?source=rss-0b668e3fef6c------2</link>
            <guid isPermaLink="false">https://medium.com/p/a9c9d0bc58b5</guid>
            <category><![CDATA[entropy]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Rorschach]]></dc:creator>
            <pubDate>Sun, 25 Aug 2024 15:04:42 GMT</pubDate>
            <atom:updated>2024-08-25T15:07:29.304Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*UNbcnRTQ_jFc0LbEnFvy3Q.png" /></figure><p>I was watching “Key Equations Behind Probability” and wanted to prepare an article on entropy in ML models.</p><p>Have a good read!</p><p><strong>What is entropy, and why is it important for ML?</strong></p><p>Entropy is a concept from thermodynamics and information theory that represents a measure of uncertainty, randomness, or chaos in a system.</p><p>In thermodynamics, entropy refers to how much of the energy in a system cannot be converted into work. In other words, it is a measure of the disorder in the system. In an isolated system, entropy tends to increase, which means that the disorder of the universe increases over time.</p><p><em>This is known as the second law of thermodynamics.</em></p><p>In information theory, entropy measures the uncertainty or information in a data set. This concept, defined by Claude Shannon, is used to determine how much information an information source contains and how much uncertainty that information contains.</p><p><em>High entropy indicates high uncertainty in the system, while low entropy indicates low uncertainty.</em></p><p>Entropy plays a central role in machine learning as a fundamental concept in information theory that quantifies uncertainty and variability in data. Applications of entropy in machine learning include classification, regression, and optimization of learning algorithms. For example, the maximum entropy principle is widely used in statistical inference and probabilistic reasoning, allowing for the construction of models that account for incomplete knowledge of probability distributions (Chen &amp; Príncipe, 2012).</p><p>This principle is particularly relevant in scenarios where the goal is to maximize the information gained from the data while minimizing assumptions about the underlying distribution, thus enhancing the robustness and interpretability of machine learning models (Yang et al., 2022).</p><p>In recent years, the integration of entropy-based methods into machine learning frameworks has gained traction, particularly in the context of improving classification accuracy and model performance. For example, the <strong>Extreme Entropy Machine</strong> (EEM) and its variants have been developed to provide robust classification capabilities by exploiting the principles of entropy (Czarnecki &amp; Tabor, 2015).</p><p>These models use entropy as a measure of uncertainty, which allows them to adapt to different data distributions and improve their predictive capabilities. Furthermore, the application of Shannon entropy in decision tree algorithms exemplifies how entropy can be effectively used to guide splitting decisions at each node, thereby optimizing the overall structure of the decision tree (Zhang, 2023).</p><p>The concept of entropy is not limited to classical machine learning paradigms; it has also found applications in quantum machine learning. For example, quantum cross entropy has been proposed as a means to improve quantum data compression protocols and the efficiency of quantum machine learning algorithms (Shangnan, 2021).</p><p>This intersection of quantum mechanics and machine learning underscores the versatility of entropy as a tool for understanding and optimizing complex systems, whether classical or quantum in nature. The exploration of quantum cross-entropy in the context of maximum likelihood further illustrates the potential for entropy-based approaches to inform the development of advanced machine learning techniques (Shangnan &amp; Wang, 2021).</p><p>In addition to its theoretical implications, entropy has practical applications in several domains, including healthcare and diagnostics. For example, the use of Shannon entropy in medical diagnostics has been shown to increase the efficiency of detecting conditions such as epilepsy by transforming large EEG datasets into a format suitable for machine learning analysis (Broberg &amp; Lewis, 2014).</p><p>This application highlights the utility of entropy as a metric for quantifying uncertainty in medical data, thereby facilitating more accurate diagnoses and treatment plans. Furthermore, the integration of entropy-based methods in machine fault diagnosis demonstrates the adaptability of these techniques across domains, demonstrating their relevance in both engineering and healthcare contexts (Zhang et al., 2023).</p><p>The relationship between entropy and variability is another critical aspect of its application in machine learning. Entropy optimization can be used to capture the underlying structure of data, especially in supervised learning scenarios where the goal is to minimize the entropy of the error between model predictions and actual results (Chen et al., 2016). This approach not only improves the accuracy of predictions but also contributes to the overall robustness of the learning process. Furthermore, the concept of structural entropy extends the traditional understanding of Shannon entropy and provides a framework for analyzing more complex, structured systems (Li, 2020). This extension is particularly valuable in scenarios where data exhibit intricate relationships that cannot be adequately captured by traditional entropy measures.</p><p>Exploring entropy in the context of unbalanced data sets is another important area of research. Traditional machine learning classifiers often struggle with unbalanced data, leading to biased results that favor the majority class. The development of instance-based entropy fuzzy support vector machines (IEFSVM) addresses this challenge by using entropy information from k-nearest neighbors to prioritize the importance of each sample, thereby improving classification performance on imbalanced datasets (Cho et al., 2019). This innovative approach highlights the adaptability of entropy-based methods to address specific challenges in the machine learning landscape.</p><p>Furthermore, the application of entropy in adaptive signal processing and machine learning highlights its role in improving the efficiency and effectiveness of learning algorithms. By incorporating entropy analysis into the learning process, researchers have demonstrated improvements in model performance, particularly in dynamic environments where data characteristics may change over time (Ogunfunmi, 2022). This adaptability is critical for developing machine learning systems that can respond to real-world challenges and evolving data landscapes.</p><p>In summary, the multifaceted role of entropy in machine learning encompasses theoretical foundations, practical applications, and innovative methods that enhance the robustness and interpretability of models. From its use in classification and regression to its applications in quantum machine learning and healthcare diagnostics, entropy serves as an essential tool for understanding and optimizing complex systems. As research continues to evolve, the integration of entropy-based approaches is likely to lead to further advances in machine learning, paving the way for more effective and adaptive algorithms.</p><p><strong>References</strong>:</p><p>Broberg, R. and Lewis, R. (2014). Classification of epileptoid oscillations in eeg using Shannon’s entropy amplitude probability distribution., 247–252. <a href="https://doi.org/10.1007/978-3-319-11988-5_23">https://doi.org/10.1007/978-3-319-11988-5_23</a> Chen, B., and Príncipe, J. (2012).</p><p>Some further results on the minimum error entropy estimation. Entropy, 14(5), 966–977. <a href="https://doi.org/10.3390/e14050966">https://doi.org/10.3390/e14050966</a> Chen, B., Wang, J., Zhao, H., &amp; Prı́ncipe, J. (2016).</p><p>Insights into entropy as a measure of multivariate variability. Entropy, 18(5), 196. <a href="https://doi.org/10.3390/e18050196">https://doi.org/10.3390/e18050196</a> Cho, P., Lee, M., &amp; Chang, W. (2019).</p><p>Instance-based entropy fuzzy support vector machine for imbalanced data. Pattern Analysis and Applications, 23(3), 1183–1202. <a href="https://doi.org/10.1007/s10044-019-00851-x">https://doi.org/10.1007/s10044-019-00851-x</a> Czarnecki, W., and Tabor, J. (2015).</p><p>Extreme entropy machines: robust information theoretic classification. Pattern Analysis and Applications, 20(2), 383–400. <a href="https://doi.org/10.1007/s10044-015-0497-8">https://doi.org/10.1007/s10044-015-0497-8</a> Li, A. (2020).</p><p>Structural information learning machinery: learning from observing, associating, optimizing, decoding, and abstracting. <a href="https://doi.org/10.48550/arxiv.2001.09637">https://doi.org/10.48550/arxiv.2001.09637</a> Ogunfunmi, T. (2022).</p><p>Adaptive signal processing and machine learning using entropy and information theory. Entropy, 24(10), 1430. <a href="https://doi.org/10.3390/e24101430">https://doi.org/10.3390/e24101430</a> Shangnan, Z. (2021).</p><p>Quantum data compression and quantum cross-entropy. <a href="https://doi.org/10.48550/arxiv.2106.13823">https://doi.org/10.48550/arxiv.2106.13823</a> Shangnan, Z. and Wang, Y. (2021).</p><p>Quantum cross entropy and maximum likelihood principle. <a href="https://doi.org/10.48550/arxiv.2102.11887">https://doi.org/10.48550/arxiv.2102.11887</a> Yang, S., Tan, J., &amp; Chen, B. (2022).</p><p>Robust spike-based continual meta-learning improved by restricted minimum error entropy criterion. Entropy, 24(4), 455. <a href="https://doi.org/10.3390/e24040455">https://doi.org/10.3390/e24040455</a> Zhang, L., Zhang, Y., &amp; Li, G. (2023).</p><p>Fault-diagnosis method for rotating machinery based on svmd entropy and machine learning. Algorithms, 16(6), 304. <a href="https://doi.org/10.3390/a16060304">https://doi.org/10.3390/a16060304</a> zhang, A. (2023).</p><p>A new information theory of certainty for machine learning. <a href="https://doi.org/10.21203/rs.3.rs-2865625/v1">https://doi.org/10.21203/rs.3.rs-2865625/v1</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=a9c9d0bc58b5" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Relationship Between Mathematics and Blockchain]]></title>
            <link>https://medium.com/@0x_Rorschach/the-relationship-between-mathematics-and-blockchain-0aeb1f78a92e?source=rss-0b668e3fef6c------2</link>
            <guid isPermaLink="false">https://medium.com/p/0aeb1f78a92e</guid>
            <category><![CDATA[science]]></category>
            <category><![CDATA[cryptocurrency]]></category>
            <category><![CDATA[mathematics]]></category>
            <category><![CDATA[blockchain]]></category>
            <dc:creator><![CDATA[Rorschach]]></dc:creator>
            <pubDate>Thu, 22 Aug 2024 09:30:00 GMT</pubDate>
            <atom:updated>2024-08-22T09:30:00.455Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*CABCBGb_LaSWW1vapQTFQw.png" /></figure><p><strong>1. Introduction</strong></p><p>The purpose of this article is to highlight the essential role of mathematics in both understanding and building blockchain technology. While the topic is extensive and complex, it’s important to recognize that mathematics is not just a collection of symbols and formulas used for computation. Instead, mathematics is a discipline that explores the relationships between different structures, frameworks, and systems. These relationships allow us to describe phenomena with greater precision, uncover hidden properties, and even drive new scientific breakthroughs. Given its fundamental role in analyzing complex systems, mathematics is essential to understanding the intricacies of blockchain.</p><p>This essay consists of three main parts, which address different aspects of the mathematical nature of blockchain. The first part provides a brief overview of the mathematical background of blockchain. There will be no formulas or calculations in this part, as its sole purpose is to provide a rationale for the need for mathematics to create and understand something as complex and sophisticated as blockchain. The second part will detail the role of game theory in the creation of the blockchain. The last part will cover more technical concepts, describing blockchain as a computer simulation of random processes and the idea of entropy and randomness in the blockchain.(1)</p><p><strong>2. Cryptography</strong><br>At its core, the relationship between mathematics and blockchain begins with a branch of study known as cryptography. Cryptography is how we protect information from being altered, intercepted, and more. While this field was once focused on protecting classified and/or sensitive government secrets, it has since become more widespread as individuals and businesses seek ways to protect their own personal information. Cryptography is responsible for encrypting web browsing sessions, initiating chat encryption, providing us with secure and personalized email interactions, and much more. In essence, cryptography lives at the intersection of information technology, mathematics, and computer science. For this reason, it’s also fundamental to the development and function of blockchain technology. (2)</p><p>Cryptography relies on mathematical algorithms to secure information and communication. There are a few fundamental “tasks” at the core of this field. First, we try to create methods for encrypting and decrypting messages. An encrypted message appears meaningless to anyone who might intercept it, thus protecting important information from prying eyes. Then there’s the need to create mathematical methods for authenticating and verifying message integrity. This helps two parties know that a message sent by their correspondent is indeed from that real correspondent and that it hasn’t been altered in any way. Finally, there are ways to sign documents, use message hashes to confirm the sender, and protect against data forgery. These cornerstones of encryption, authentication, and integrity help lay the foundation for blockchain technology. (3)</p><p><strong>3. Consensus Algorithms</strong><br>From this described architecture, we introduce the topic of the paper: the mathematical principles that underlie blockchain paradigms. At the foundation of each is a consensus algorithm, which ensures that the connected network of miners (or nodes) agree on the correct state of the ledger. By breaking down the discussion of blockchain consensus algorithms into their mathematical properties, we assert the fiduciary requirements of different types of systems and discuss how these mathematical and consensus-based blockchains have broad implications for trustless economies and their further potential applications. (4)</p><p>Consensus algorithms are the foundation of distributed systems, providing solutions to classically difficult problems in distributed computing. Within a blockchain, the consensus algorithm ensures agreement and proves that the chosen version of the blockchain-which forms the set of confirmed transactions-is the “correct” version according to the probability distribution of values voted on by the network partners. In addition, within a blockchain, the consensus algorithm—when voting on the correct version—also elects one or a few governments in the network. The process by which the network operates is simplified down to the lowest level, so that the entirety of operations can be governed by some output function or the output of some oracle. In this way, the complex system of our world is built on the back of a mathematical consensus. (4)</p><p><strong>4. Game Theory</strong><br>Mathematics is the mathematical formalization of decision-making in an adversarial and strategic situation involving two or more decision-makers. Depending on the scale, there are zero-sum games, constant-sum games, and non-constant-sum games; depending on the time, there are simultaneous games and sequential games; depending on the information, there are complete information games and incomplete information games; depending on whether the opponent is known, there are two-person games and two-person games; depending on whether the game object is different, there are different rights. In addition to games for equal rights, there are also games for unequal rights. Game theory is widely used in economics, military science, political science, biology and other fields. Researchers in recent years have begun to apply ideas in game theory to various problems in blockchain systems, mainly embodied in mathematics. (5)</p><p>Although game theory is widely applied in the aforementioned areas, researchers tend to focus on different aspects in different studies. For clarity in this article, I categorize the application of game theory in blockchain into three broad areas. The first category emphasizes mathematical models that are commonly used in the study of blockchain mechanism design. This includes work on exponential mechanisms and mining pool structures, where game theory is applied alongside probability distributions and game models. The design of game-theoretic mechanisms for both PoW (Proof of Work) and PoS (Proof of Stake) systems has received considerable attention in recent years. In particular, there has been a focus on designing secure game-theoretic mechanisms for these systems to mitigate the risk of attacks by key participants. With this in mind, I explore the particular mechanism designs that are tailored to specific blockchain systems.(5)</p><p><strong>5. Linear Algebra</strong><br>Blockchains offer solutions to several computational challenges, often leveraging cryptographic methods grounded in abstract algebra. Additionally, concepts from linear algebra, such as matrix operations and vector spaces, play a significant role. The use of matrix operations is particularly relevant due to their capacity to model the data within these systems. Most existing literature primarily focuses on pairwise trusts or raters. However, blockchain networks are designed to accommodate a large number of participants, making the representation of the transaction graph through matrices essential. Moreover, symmetric matrices are commonly employed in cryptographic techniques. For instance, some research has explored modeling the transaction graph using quadratic forms and integral solutions, providing an alternative approach to understanding these systems. (7)</p><p>In recent years, matrix operations have proven to be a valuable technique for improving computational efficiency. For example, matrix multiplication has been used to speed up tasks such as magazine ranking, community detection, and security-based broadcasting problems. A key operation within matrix multiplication is the dot product of vectors, which measures the relevance between two vectors and serves as an important metric in fields such as information retrieval and data mining. A blockchain can be represented as a directed graph, where nodes correspond to accounts and edges represent the movement of money between them. Transaction graphs are typically sparse, with most computed features being zero. The singular vectors derived from transaction graph data often do not directly distinguish between anomalous and normal transaction histories in an interpretable way. In addition, singular vectors in this context are used for different purposes than the singular spectra of a standard adjacency matrix.(8)</p><p><strong>6. Probability and Statistics</strong><br>Probability and statistics are introduced when discussing the operations and events associated with blockchain. Discussing the reliability, trust, and decentralization of consensus mechanisms leads us to probability theory, since any declarative statement about these unrealized future events is essentially a Bayesian or subjective statement about an instantiated process. Probability describes how such a process maps potential decisions or actual outcomes in relation to many other interconnected and interdependent system participants. The use of confidence intervals, standard deviations, and other statistical anxieties are artifacts that numerators within classical probability theory use to facilitate greater confidence under uncertainty. (9)</p><p>Blockchain, by definition, is an implementation of a peer-to-peer, legitimately decentralized network without the requirements of a mutually recognized third-party intermediary. However, there are still many coordinated efforts as well as mechanical and systematic methods to assess risk, or lack thereof, in the system. The discussion here focuses on the use of probability theory and statistical analysis in the context of blockchain data validation, as well as risk mitigation measures for system security against various types of attacks. (10)</p><p>Transactions in a decentralized system such as the blockchain are validated by a method called consensus. The concept of blockchain originated with the bitcoin cryptocurrency, which has a proof-of-work consensus mechanism. The basic functions of a consensus mechanism are to prevent double spending and to select the next block to be added to the blockchain. However, many distributed ledgers, especially permissioned blockchain systems, use consensus mechanisms to validate transactions and prevent attacks such as DDoS, Eclipse and Sybil attacks. The consensus mechanisms are proven to validate transactions and mitigate security risks. When transactions are validated, they are stored in the system in the form of blocks. With these consensus mechanisms, there is a leader election process to select the next miner/verifier to participate in the block addition process. In this paper, we provide a detailed pruning of each consensus mechanism, issues related to them, and their pros and cons. In addition, we discuss issues related to the consensus mechanism and future research directions. Accordingly, we have proposed various research solutions. The goal of the consensus mechanism is to establish an agreement among the network participants and provide better synchronization of the blockchain system. (11)</p><p><strong>7. Elliptic curve mathematics</strong><br>Elliptic curve mathematics provides an inherent security feature that is largely responsible for the success of blockchains today. This elliptic curve cryptography (ECC) is also the underlying algorithm for proving ownership of an account, confirming an outgoing transaction, and more. In this section, we’ll look at the basics of elliptic curves and how they are used in cryptocurrency and blockchain technology. (12)</p><p>A group structure can be formed in which an elliptic curve and points within it can be added together for an operation such as 2⋅P = E(P+P), where P is a point and E is the elliptic curve equation. This group is used to transfer coin amounts, and the owner’s private key can be used directly to make a transfer or receive coins, depending on the system. The signature is used to prove ownership of coins, where the owner prepares a message, signs it using the private key, and then gives the message and signature to the other party. Depending on the system, by establishing a legitimate link between a public key and an identity, one can securely sign messages. (13)</p><p>Because blockchain systems are peer-to-peer networks, it is impossible to know which node a user is actually communicating with. In addition, public keys and identities are directly linked, so one can sign messages and send them to the intended recipient with the guaranteed knowledge that only that recipient will be able to decrypt the message because only that recipient has the private key. In any scenario, including blockchains, one can perform these actions in a fast, lightweight process based on the group of points on an elliptic curve. (14)</p><p><strong>8. Algebra and Number Theory</strong><br>Number theory underlies almost all aspects of modern cryptography. The basic operations that form cryptographic schemes—secure signatures, private communications, and secure computations—are based on the mathematical properties of prime numbers, lattices, and abelian groups. Continuing the thread of number theory, addition and multiplication on the set of integers can be thought of as two kinds of multiplication operations defined on secret sharing polynomials. In group and ring theory, the study of abelian groups and fields forms the basis for public-key statistical zero-knowledge proofs used in blockchains, such as σ-protocols and bulletproofs. The hidden community proof systems of Micali et al. use non-abelian groups. (15)</p><p>Algebra contains a vast toolbox that we constantly invoke when tinkering with, designing, and analyzing Byzantine fault-tolerant consensus protocols. We rely on symmetry groups of graphs to create honest leadership election systems in PoET and Hashgraph, use isomorphisms to automatically change details of a permissioned BFT system without compromising security, and add decades-old college-level algorithms for factoring random bipartite graphs (a Higig graph) to improve the performance of BFT consensus protocols. Importantly, a rudimentary understanding of this area of mathematics is common among computer scientists. The mathematical techniques in this area are old and well-tested, and researchers and protocols are constantly making more stringent cryptographic security requirements and striving to develop efficient proofs that these cryptographic assumptions hold. Thus, no new mathematical breakthrough is likely to provide surprising new algorithmic primitives in the near future, but it remains quite possible on the horizon to elucidate the properties of prime numbers that make it difficult to prove the security of RSA or to find more interesting algorithmic problems that appear to have the intractability of integer factorization. (16)</p><p><strong>9. Conclusion</strong><br>In this work, I have explored the deep and evolving connection between mathematics and blockchain technology. It’s clear that different mathematical concepts are fundamental to both the functionality and security of blockchain systems. Algebra and number theory support the wallets, hash functions, and cryptographic elements that are critical to these protocols. Without advanced techniques such as pairings or lattices, blockchain systems would be very different from what we know today. In addition, fields like functional analysis and abstract probability theory play a key role in assessing incentives and rationality within blockchain consensus algorithms.</p><p>As I have demonstrated, the relationship between mathematics and blockchain is not only significant but also expanding. With the ongoing proliferation of cryptosystems and blockchain platforms, this relationship is poised to deepen further. For instance, Ethereum 2.0’s reliance on a PoS consensus algorithm highlights the integration of mathematical finance theories into blockchain design. This interplay between blockchain and mathematics is bidirectional; the emergence of cryptocurrencies is driving new theoretical advancements in areas like dynamic mechanism design.</p><p><strong>References:</strong></p><p>1. Song L, Luo Y, Chang Z, Jin C, et al. Blockchain adoption in agricultural supply chain for better sustainability: A game theory perspective. Sustainability. 2022. <a href="https://www.mdpi.com/2071-1050/14/3/1470/pdf">mdpi.com</a></p><p>2. Guo H., Yu X. A survey on blockchain technology and its security. Blockchain: research and applications. 2022. <a href="https://www.sciencedirect.com/science/article/pii/S2096720922000070">sciencedirect.com</a></p><p>3. ALRikabi HT, Hazim HT. Enhanced data security of communication system using combined encryption and steganography. iJIM. 2021. <a href="https://lib.uowasit.edu.iq/books/Enhanced_Data_Security_of_Communication.pdf">uowasit.edu.iq</a></p><p>4. Bamakan SMH, Motavali A, Bondarti AB. A survey of blockchain consensus algorithm performance evaluation criteria. Expert Systems with Applications, 2020. <a href="https://www.sciencedirect.com/science/article/pii/S0957417420302098">[HTML]</a></p><p>5. Maschler M, Zamir S, and Solan E. Game Theory, 2020. <a href="https://books.google.com/books?hl=en&amp;lr=&amp;id=o6UlEAAAQBAJ&amp;oi=fnd&amp;pg=PR14&amp;dq=Game+theory+is+widely+used+in+economics,+military,+political+science,+and+biology.&amp;ots=HlzTLnMwih&amp;sig=LK7AkngCnpotv7pE53oxEfmuuI4">[HTML]</a></p><p>6. Ferdowsi A, Eldosouky A, and Saad W. Interdependence-aware game-theoretic framework for secure intelligent transportation systems. IEEE Internet of Things Journal. 2020 Sep 1;8(22):16395–405. <a href="https://arxiv.org/pdf/2007.05904">[PDF]</a></p><p>7. Lipton A., Treccani A. Blockchain and distributed ledgers: mathematics, technology, and economics, 2021. <a href="https://www.worldscientific.com/doi/pdf/10.1142/9789811221538_0001">worldscientific.com</a></p><p>8. Frikha T, Chaabane F, Aouinti N, Cheikhrouhou O, Ben Amor N, Kerrouche A. Implementation of blockchain consensus algorithm on embedded architecture. Security and Communication Networks, 2021(1):9918697. <a href="https://onlinelibrary.wiley.com/doi/pdf/10.1155/2021/9918697">wiley.com</a></p><p>9. Yan L, Yin-He S, Qian Y, Zhi-Yu S, Chun-Zi W, and Zi-Yun L. Method of reaching consensus on probability of food safety based on the integration of finite credible data on block chain. IEEE access. 2021 Aug 26;9:123764–76. <a href="https://ieeexplore.ieee.org/iel7/6287639/6514899/09523863.pdf">ieee.org</a></p><p>10. Hassija V, Bansal G, Chamola V, Kumar N, and Guizani M. Secure lending: Blockchain and prospect theory-based decentralized credit scoring model. IEEE Transactions on Network Science and Engineering. 2020 Mar 23;7(4):2566–75. <a href="https://ieeexplore.ieee.org/abstract/document/9044411/">[HTML]</a></p><p>11. Hashemi Joo M., Nishikawa Y., and Dandapani K. Cryptocurrency is a successful application of blockchain technology. Managerial Finance, 2020. <a href="https://www.emerald.com/insight/content/doi/10.1108/MF-09-2018-0451/full/html">[HTML]</a></p><p>12. Woda M, Huzaini Z. A proposal to use elliptical curves to secure the block in e-voting system based on blockchain mechanism. International Conference on Dependability and Complex Systems 2021, May 27 (pp. 466–476). Cham: Springer International Publishing. <a href="https://www.researchgate.net/profile/Marek-Woda/publication/351897948_A_Proposal_to_Use_Elliptical_Curves_to_Secure_the_Block_in_E-voting_System_Based_on_Blockchain_Mechanism/links/60b490d1299bf1f6d58820d1/A-Proposal-to-Use-Elliptical-Curves-to-Secure-the-Block-in-E-voting-System-Based-on-Blockchain-Mechanism.pdf">researchgate.net</a></p><p>13. Häner T., Jaques S., Naehrig M., Roetteler M., and Soeken M. Improved quantum circuits for elliptic curve discrete logarithms. In Post-Quantum Cryptography: 11th International Conference, PQCrypto 2020, Paris, France, April 15–17, 2020, Proceedings 11 2020 (pp. 425–444). Springer International Publishing. <a href="https://arxiv.org/pdf/2001.09580">[PDF]</a></p><p>14. Carrara GR, Burle LM, Medeiros DS, de Albuquerque CV, and Mattos DM. Consistency, availability, and partition tolerance in blockchain: a survey on the consensus mechanism over peer-to-peer networking. Annals of Telecommunications. 2020 Apr;75:163–74. <a href="https://link.springer.com/article/10.1007/s12243-020-00751-w">[HTML]</a></p><p>15. Buell D. Fundamentals of Cryptography: Introducing Mathematical and Algorithmic Foundations, 2021. <a href="https://books.google.com/books?hl=en&amp;lr=&amp;id=LH8zEAAAQBAJ&amp;oi=fnd&amp;pg=PR5&amp;dq=Number+theory+underpins+nearly+all+aspects+of+modern+cryptography.&amp;ots=Db_-b1noLX&amp;sig=g4n3DUUT7p_fp-8OOOE8-XWCYZo">[HTML]</a></p><p>16. Caeli EN, Yadav A. Unplugged approaches to computational thinking: A historical perspective. TechTrends. 2020. <a href="https://www.researchgate.net/profile/Aman-Yadav-12/publication/334280681_Unplugged_Approaches_to_Computational_Thinking_a_Historical_Perspective/links/5f070927299bf188160e723e/Unplugged-Approaches-to-Computational-Thinking-a-Historical-Perspective.pdf">researchgate.net</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=0aeb1f78a92e" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The Innovation Game]]></title>
            <link>https://medium.com/@0x_Rorschach/the-innovation-game-482b8c9ed917?source=rss-0b668e3fef6c------2</link>
            <guid isPermaLink="false">https://medium.com/p/482b8c9ed917</guid>
            <category><![CDATA[science]]></category>
            <category><![CDATA[open-source]]></category>
            <category><![CDATA[innovation]]></category>
            <dc:creator><![CDATA[Rorschach]]></dc:creator>
            <pubDate>Mon, 19 Aug 2024 08:10:00 GMT</pubDate>
            <atom:updated>2024-08-19T10:01:12.476Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*VkKut7h9-6D49kkz" /></figure><blockquote>For a few days, I have been preparing you to better understand a project.</blockquote><h3>The Innovation Game</h3><p>The development of computational methods is essential to the advancement of data-driven sciences. Algorithms play a critical role in advancing our understanding of and solving complex problems in a variety of fields. However, the traditional approach to developing these methods often limits collaboration and stifles innovation. Closed systems and proprietary frameworks create barriers that make it difficult for new ideas to emerge and gain traction.</p><p>To address these challenges, a new framework is needed-one that encourages open collaboration and values contributions based on value. Such a framework would allow a wide range of contributors to participate in the creation and refinement of algorithms, thereby advancing scientific development. By creating a transparent environment where the value of computational methods is determined by their effectiveness and adoption, we can create a more inclusive and innovative culture.</p><p>This approach not only enhances the scientific community’s ability to solve complex challenges but also creates opportunities to attract private investment. By integrating open source principles with a market-driven approach, we can accelerate the development of key computational methods and ensure that innovation is accessible to all.</p><p>From this point of view, I would like to introduce you a project that can achieve these goals</p><blockquote>“The Innovation Game”<br> <br>The Innovation Game (TIG) is a new market-based system designed to accelerate the creation and improvement of computational methods essential to data-driven scientific research. TIG provides a platform where these methods can be developed, shared, and rewarded, facilitating a collaborative environment that drives progress. This system not only supports the creation of new algorithms but also encourages their continuous improvement through real-world application, making it easier and faster to address scientific challenges.</blockquote><p>The Innovation Game (TIG) is designed to accelerate the development of computational methods specifically aimed at solving “<strong>Asymmetric Problems</strong>”. These are problems where finding a solution requires a significant amount of computation, but once the solution is proposed, it is easy to verify that it is correct. You can think of it like solving a puzzle: it may take a long time to put the pieces together, but once the puzzle is complete, it’s obvious that the solution is correct.</p><p>A prominent category of these problems are NP-complete problems. These problems are notable because, while they are easy to verify once a solution is found, finding that solution is generally considered infeasible in polynomial time. A classic example of an NP-complete problem is the Hamiltonian cycle problem, which involves determining whether a given graph contains a cycle that visits each vertex exactly once. Another well-known NP-complete problem is the Boolean Satisfiability Problem (SAT), where the challenge is to decide whether there exists an assignment of variables that satisfies a given Boolean formula.</p><p>In addition to these classic examples, another important type of asymmetric problem is found in the field of artificial intelligence, specifically in the training of machine learning models. Building a model based on a data set requires significant computational resources, especially as the complexity and size of the data increases. However, once trained, the model’s performance on specific tasks, such as image recognition or natural language processing, can be quickly evaluated. In scientific research, machine learning techniques enable the processing of large data sets, which in turn facilitates the formulation and testing of complex hypotheses. This capability is increasingly critical in our data-centric world, where the ability to effectively handle large amounts of data is a cornerstone of modern scientific inquiry.</p><p>By focusing on the development and optimization of algorithms to solve such asymmetric problems, TIG aims to create a marketplace where these computational methods can be refined, shared, and adopted across fields. This approach not only drives innovation in algorithm development but also ensures that the solutions generated are practical and widely applicable, ultimately benefiting the broader scientific and engineering communities.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*uP-_VP5lSeNV6XLTh9BAXQ.png" /></figure><p>1. Computational science is a subset of science; a subset of problems in computational science are asymmetric (meaning that solutions are difficult to find but easy to verify — see above); 2. Scientific research can be broadly divided into fundamental (pre-commercial) and applied (commercial); 3. Value capture combined with price discovery can address issues (a) and (b), above, for research in asymmetric computational problems</p><p><strong>The importance of Fundamental and Applied research</strong></p><p>Scientific research can be broadly divided into fundamental and applied science. Fundamental research is concerned with expanding our understanding of natural phenomena. It often leads to discoveries that lay the foundation for new technologies. For example, the Internet, a vital tool for global communication, grew out of fundamental research in computer science. Similarly, GPS technology, which has revolutionized navigation and location services worldwide, is the result of fundamental research in physics. Another example is the discovery of the structure of DNA, a breakthrough in biology that laid the foundation for modern genetics and biotechnology.</p><p>Applied science, on the other hand, focuses on translating this knowledge into practical solutions that improve our daily lives. These advances not only improve the quality of life but also drive economic growth. The profits generated by these innovations often feed back into more fundamental research, creating a cycle of continuous improvement and discovery.</p><p>However, there are significant economic challenges associated with this type of research:</p><p>1. Commercial (applied) research: While private investment is typically available, it usually comes with the condition that research results, including data, are proprietary. This exclusivity can discourage open collaboration, which is essential for faster scientific progress.</p><p>2. Pre-commercial (fundamental) research: This type of research is often non-proprietary, meaning that its results are openly accessible. However, because there is no immediate financial return, private investment is scarce, leading to underfunding and inefficient allocation of resources.</p><p><strong>The Role of Open Source in Accelerating Science</strong></p><p>To overcome these challenges, there is a need for a “<strong>framework that supports open, collaborative development of computational methods.</strong>”. Open source development naturally fits this role, as it encourages sharing and collaboration. However, in his essay “The Magic Cauldron,” Eric Raymond points out that while open source development has many strengths, it also has limitations, particularly for projects that rely heavily on complex code, such as those in computational science.</p><p>Raymond suggests that for open source projects to be viable in these areas, they must include:</p><p>A value capture mechanism: A way to capture the economic value generated by the project so that contributors can be rewarded for their efforts.</p><p>A price discovery mechanism: A method for determining the value of individual contributions so that rewards can be fairly distributed according to the impact of each contribution.</p><p><strong>Challenges of Open Collaborative Development</strong></p><p>One of the major challenges in open collaborative development, especially in areas such as machine learning, is data sharing. In many jurisdictions, data sets are not considered intellectual property, which means they cannot be protected by traditional copyright laws. This creates a problem where open source projects that share data are vulnerable to “free riding” by closed source projects. These closed projects can use the data without contributing anything in return, undermining the collaborative spirit of open source initiatives.</p><p>To address this, <strong>The Innovation Game</strong> aims to incentivize data sharing through a special licensing strategy. This strategy is designed to encourage reciprocal data sharing, ensuring that open source projects remain sustainable and that contributors are fairly compensated for their work.</p><h4><strong>Synthetic Market for Price Discovery in Computational Methods</strong></h4><p>The Innovation Game addresses key challenges in computational science by creating a structured approach to price discovery and value capture. These challenges often revolve around resource allocation and the proprietary nature of research results, which can inhibit collaboration and innovation. To address these issues, The Innovation Game introduces a synthetic market — an artificially constructed marketplace that is critical for determining the value of computational methods (referred to as “methods”).<br> <br><strong> What is a synthetic market?</strong></p><p>A synthetic market is not a naturally occurring one where supply and demand operate as usual. Instead, it is a deliberately created environment designed to foster the development and optimization of methods for solving specific problems in computational science, referred to as “Challenges”.</p><p>Agents referred to as benchmarkers are responsible for generating demand in this artificial market. These agents simulate demand by using different methods to solve instances of the Challenges. Their role is similar to that of customers in a traditional market, but with a specific focus on optimizing performance rather than simply purchasing a product. As benchmarkers strive to find the most efficient methodologies, they inadvertently help determine the true value of those methodologies-this process is known as price discovery.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*hl9HjL0CS_0y02rnQS7KiQ.png" /></figure><p><strong>How Does Price Discovery Work?</strong></p><p>Price discovery in The Innovation Game is a process by which the value of different Methods is revealed through their adoption and performance in the marketplace. When benchmarkers use a method to solve a challenge and find it effective, they contribute to the perceived value of the method. This is similar to how, in traditional markets, a product that meets customer needs and expectations gains value through increased demand.</p><p>Innovators who develop and submit these Methods are rewarded based on how widely their Methods are adopted by benchmarkers. The more a method is used, the higher its value in the synthetic market, and the more tokens the innovator earns. This mechanism encourages continuous improvement and innovation, as innovators are incentivized to develop better and more efficient methods.</p><p><strong>Ensuring Value Capture</strong></p><p>To ensure that the value generated by these Methods is captured and benefits the entire ecosystem, The Innovation Game secures the intellectual property (IP) rights to these Methods through the TIG Foundation. The Foundation licenses these methods both for open collaboration (under a TIG Open Data license) and for commercial use (under a TIG Commercial license). This dual licensing strategy ensures that, while the methods are available for collaborative development, they can also generate revenue through commercial applications.</p><p><em>Think of the synthetic market in The Innovation Game as a kind of “laboratory marketplace” where algorithms are tested, refined, and evaluated based on their real-world performance. Just as scientists in a laboratory experiment to find the most effective solutions, benchmarkers in this marketplace experiment with different methods to find the most efficient solutions to computational challenges. The value of each method is determined by how well it performs in these “experiments,” and innovators are rewarded accordingly, much like scientists who gain recognition and funding for successful experiments. This synthetic market thus acts as both a testing ground and a scoring system, driving the continuous improvement of computational methods in ways that benefit the entire scientific community.</em></p><h4><strong><em>Price Discovery in the Innovation Game</em></strong></h4><p>What is price discovery?</p><p>Price discovery is the process of determining the value of something through the interactions of buyers and sellers in a market. In The Innovation Game, price discovery is about figuring out how much a particular algorithm, or “method,” is worth based on how well it performs in solving specific computational problems, known as “challenges.</p><p>The Role of Proof of Work (PoW)</p><p>The Innovation Game uses a system called “Proof of Work” (PoW) to aid in price discovery. PoW is a method in which participants (called benchmarkers) solve complex problems, and the completion of these problems can be easily verified by others. This method is borrowed from digital currencies such as bitcoin, where it helps maintain the security of the system by ensuring that solving problems requires significant effort.</p><p>Challenges with Traditional PoW</p><p>In traditional PoW systems, the tasks given to participants are deliberately made difficult and time-consuming, but there’s little room for improvement through better algorithms. This is because if one participant could drastically improve its efficiency, it could dominate the market, leading to centralization and reducing the fairness and security of the system.</p><blockquote><strong>Optimizable Proof of Work (OPOW)</strong></blockquote><blockquote>To address this problem, The Innovation Game introduces a concept called “Optimizable Proof of Work” (OPOW). Unlike traditional PoW, OPOW involves problems that can be optimized, meaning that with better algorithms, these problems can be solved more efficiently. However, The Innovation Game carefully selects these problems (challenges) to ensure that optimizing one doesn’t lead to similar improvements in others. This prevents any one participant from gaining an overwhelming advantage.</blockquote><p>Maintaining Competition and Fairness</p><p>In a game theory context, The Innovation Game must prevent what’s called “<strong>dominant strategy&quot;—whereone</strong> player finds a method so superior that it outcompetes all others, leading to monopolization. To avoid this, The Innovation Game ensures that while participants can improve their methods, the benefits of those improvements are spread across different Challenges.</p><p><em>For better understanding, I will describe this with examples:</em><br> <br><strong>The Rowing Boat</strong>: Imagine a rowing boat where each rower represents a different challenge in The Innovation Game. Just as a rowing team must synchronize their strokes to move the boat efficiently, benchmarkers in TIG must balance their performance across Challenges to maximize their rewards. Even if one rower (or method) is significantly stronger, the overall speed of the boat (or reward) is only as fast as how well all the rowers cooperate. This ensures that the competition remains fair and that no single method dominates the market.</p><p><em>I believe giving another example of this would increase understanding.</em></p><p><strong>Relay Race</strong>: Consider a relay race where each runner must complete a portion of the race before passing the baton to the next runner. In TIG, each challenge resembles a section of the race, and the innovator uses the runner as their method. Just as the fastest relay team isn’t just the one with the best sprinter, but the team that performs well across all segments, the most successful benchmarkers in TIG are those who perform consistently well across all Challenges. This setup ensures that rewards are based on overall performance, not just excellence in one area.</p><h4><strong>Ensuring Value Capture</strong></h4><p>To ensure that the value generated by these methods is captured and benefits the entire ecosystem, The Innovation Game secures the intellectual property (IP) rights to these methods through the TIG Foundation. The Foundation licenses these methods both for open collaboration (under a TIG Open Data license) and for commercial use (under a TIG Commercial license). This dual licensing strategy ensures that, while the methods are available for collaborative development, they can also generate revenue through commercial applications.</p><p><strong>Value Capture and Intellectual Property in the Innovation Game</strong></p><p>Sustainable Value of TIG Tokens</p><p>In order for The Innovation Game (TIG) to remain sustainable, it’s essential that the TIG token represent a tangible asset from which revenue can be derived. The core value of these tokens is tied to the intellectual property (IP) generated by the game. Submissions to TIG are protected by copyright, and some may even involve patentable inventions. The TIG Foundation secures and manages these IP rights, ensuring that they can generate revenue that underpins the value of the TIG tokens.</p><p>Role of Benchmarkers in Capturing Value</p><p>Benchmarkers play a critical role by signaling which methods are improvements over existing methods. This feedback helps determine whether to pursue patent applications for the underlying inventions. Even if a patent is not granted, the optimized code itself has significant value, especially if it’s difficult to replicate without infringing copyright or losing the inherent benefits of the code.</p><h4><strong>Licensing Framework in The Innovation Game</strong></h4><p>In The Innovation Game (TIG), various types of Intellectual Property (IP) are generated through the creation and optimization of computational methods, known as Methods. To manage and distribute these valuable assets, TIG provides several different licensing options. Each license serves a specific purpose and addresses the needs of different stakeholders to ensure that the IP is used effectively and fairly.<br> <br><strong> Purpose of the Licenses:</strong></p><p>The purpose of these licenses is to create a flexible system that allows innovators and benchmarkers to use, share, and improve methods while allowing the TIG Foundation to secure and commercialize the intellectual property. This system encourages collaboration, innovation, and sustainable growth of the TIG tokens by ensuring that the IP generated in the game is both protected and accessible.</p><blockquote>Here’s a brief overview of each license and its purpose:</blockquote><p><strong>1. TIG Innovator Outbound Game License:</strong></p><p>Purpose: This license allows Innovators-those who create new methods-to access and use methods that have already been submitted to TIG by others. It encourages Innovators to build on existing work, facilitating continuous improvement and innovation within the game.</p><p><strong>2. TIG Benchmarker Outbound Game License:</strong></p><p>Purpose: This license provides benchmarkers — who use the methods to solve computational challenges — with access to the submitted methods. It ensures that benchmarkers can use the best tools available to compete effectively in the game, thereby encouraging the development of more efficient solutions.</p><p><strong>3. TIG Inbound Game License:</strong><br> <br> Purpose: This license governs how new methods are submitted to TIG. It allows the TIG Foundation to secure the intellectual property rights to these innovations, ensuring that the IP generated by TIG is protected and can be effectively managed.</p><p><strong>4. TIG Open Data License:</strong></p><p>Purpose: This license is designed to promote open collaboration by requiring that the data and source code associated with the methods be shared. It promotes transparency and allows other innovators to build on the work, fostering a community of collaboration and accelerating the pace of innovation.</p><p><strong>5. TIG Commercial License:</strong></p><p>Purpose: This license provides greater flexibility for commercial use of the IP. It allows companies to use TIG’s IP without the sharing requirements of the Open Data License in exchange for a fee paid in TIG tokens. This fee helps maintain the value of the TIG tokens and supports the ongoing development and maintenance of the game.</p><p><strong>Why are these licenses important?</strong></p><p>The variety of licenses offered by The Innovation Game ensures that the intellectual property created within the game is used in a way that benefits everyone involved — innovators, benchmarkers, and the broader community. By providing structured access and defining clear rules for how the IP can be used, these licenses help maintain a balanced and fair system that encourages innovation, collaboration, and the sustainable growth of the TIG ecosystem.</p><h4><strong>Intellectual Property Management and Leverage</strong></h4><p>The TIG Foundation manages the IP secured from innovators’ submissions. The performance of the benchmarkers provides a market signal that informs patent filing decisions. Over time, the demand for TIG tokens is expected to correlate with the demand for TIG Commercial Licenses, which are required for commercial use of the IP.</p><p>Key Value Drivers for TIG’s IP</p><blockquote><em>Several factors are expected to increase the value of TIG’s IP:</em></blockquote><p><strong>Spillover effects:</strong> The diversity of challenges within TIG, spanning basic and applied research, is expected to generate spillover benefits. These benefits reflect how governments reap returns on investments in basic research through commercialization.</p><p><strong>Open collaboration:</strong> The open sharing of code and data allows innovators to build on each other’s work, a key principle of open science and a major factor in the success of open source. The more TIG contributes to the sharing of open resources, the greater its perceived value and contribution to the community.</p><p><strong>Community Building:</strong> TIG aims to create a strong community of innovators motivated by factors similar to those in academia and open source projects. Leaderboards and competitive elements will help foster this community, encourage continuous improvement, and reduce the risk of free riding on TIG’s IP.</p><blockquote>Lastly, I want to write about Game Theory and it’s relationship with The Innovation game.</blockquote><p>Understanding Game Theory</p><p>Game theory is a branch of mathematics that studies strategic interactions among rational decision-makers. It helps us understand how individuals or entities make decisions that maximize their utility while taking into account the actions and reactions of others involved in the “game”. Key concepts in game theory include Nash equilibrium, where players reach a situation where no one can improve their payoff by unilaterally changing their strategy, and dominant strategies, which are strategies that always produce a better outcome for one player, regardless of what the others do.</p><p><strong>The Innovation Game and Game Theory</strong></p><p>The Innovation Game (TIG) applies principles of game theory to design a competitive yet collaborative environment where different participants (innovators, benchmarkers, and commercial firms) interact to achieve their goals. By introducing a synthetic market and specific mechanisms for price discovery and value capture, TIG creates a strategic framework that aligns the incentives of all participants.</p><p>1. Price discovery and game theory:</p><p>In TIG, the price discovery process resembles a dynamic game in which innovators and benchmarkers interact. Innovators develop and submit methods (algorithms), while benchmarkers select and use these methods to solve computational challenges. Similar to how supply and demand determine market prices, the use and effectiveness of each method determines its value.</p><p>Strategic Interaction: Benchmarkers must decide which methods to use, balancing the need to use the most efficient tools against the competition they face from others. Innovators, on the other hand, strategize about how to create the most valuable methods that will be widely adopted and reap the highest rewards.</p><p>2. Value capture and IP management:</p><p>The value capture mechanism in TIG is also influenced by game theory. Innovators must decide whether to keep their innovations proprietary or license them under the various options provided by TIG. The TIG Foundation, acting as a central authority, manages these decisions to ensure that the system remains balanced and competitive.</p><p>Antitrust Mechanism: The introduction of an “antitrust mechanism” ensures that no single benchmarker can dominate the market, thereby preserving competition. This mechanism is a direct application of game theory, where the system is designed to prevent any player from having a dominant strategy that could lead to monopolization, which would otherwise reduce overall innovation.</p><p>3. Licensing Framework:</p><p>The different licenses offered by TIG (Innovator Outbound, Benchmarker Outbound, Inbound, Open Data, and Commercial) create a structured environment where participants can choose their strategies based on their goals. For example, an innovator may choose the TIG Open Data license if they value collaboration and broader community recognition, or they may choose the TIG Commercial license if they want to maximize revenue from their innovations.</p><p>Strategic choices: Each license represents a different strategy that participants can choose, depending on their preferences for openness versus exclusivity and collaboration versus competition. These choices are guided by the principles of game theory, where each player must consider the potential payoffs of his decisions relative to the strategies of others.</p><p>4. Spillovers and Community Building:</p><p>Spillover effects: The broader benefits of shared knowledge and innovation within TIG can be understood through the lens of game theory. As more participants contribute to the system, the collective value increases, creating a positive-sum game in which everyone benefits.</p><p>Community Building: The establishment of strong community norms and leaderboards encourages participants to adhere to shared standards and collaborate effectively, even in a competitive environment. This reflects a cooperative game where participants work together to maximize the overall benefits of the system while still competing for individual rewards.</p><p>By integrating game theory into its design, The Innovation Game creates a balanced ecosystem that fosters innovation, fair competition, and strategic collaboration. Participants are guided by a set of rules and incentives that align their actions with the broader goals of the system, ensuring that the development of computational methods remains dynamic, efficient, and sustainable. This approach not only accelerates scientific progress but also creates a robust market for intellectual property that benefits all stakeholders.</p><h3><strong>How The Innovation Game will Improve Science</strong></h3><p>The Innovation Game (TIG) is designed to make a significant contribution to the advancement of science by supporting a dynamic environment in which computational methods can be rapidly developed, tested, and refined. This will have far-reaching effects across scientific disciplines, leading to faster innovation, more efficient research, and the democratization of innovative technologies.</p><p>Here’s some examples to how TIG will improve science:</p><p><strong>1. Accelerate research and development</strong></p><p>Real-world example: Drug Discovery</p><p>In the pharmaceutical industry, the drug discovery process is notoriously slow and expensive. Computational methods, particularly those involving machine learning, can dramatically reduce the time it takes to identify promising compounds. By applying TIG’s synthetic market to drug discovery, innovators can develop and optimize algorithms for predicting molecular interactions, while benchmarkers use these algorithms to screen large libraries of compounds. The most effective algorithms, as identified by TIG’s price discovery mechanism, can then be licensed by pharmaceutical companies, accelerating the development of new drugs and bringing them to market faster.</p><p>Impact: TIG enables rapid iteration and optimization of computational models, reducing the time and cost associated with research and development. This accelerates the pace at which scientific discoveries are made and translated into real-world applications.</p><p><strong>2. Democratization of scientific tools and knowledge</strong></p><p>Real-world example: Climate modeling</p><p>Climate modeling is essential to understanding and mitigating the impacts of climate change. However, development of and access to advanced climate models is often limited to well-funded institutions. TIG can democratize access to these tools by encouraging the development of open-source climate models through its licensing framework. Innovators can contribute models that are then made available to a global community of researchers and policymakers. This open access to advanced models allows scientists in developing countries to participate in climate research and contribute valuable insights that might otherwise be overlooked.</p><p>Impact: TIG promotes the sharing and collaborative development of scientific tools, making advanced technologies more accessible to a broader audience. This democratization of knowledge leads to more inclusive and diverse scientific research.</p><p><strong>3. Enhanced collaboration and cross-disciplinary innovation</strong></p><p>Real-world example: Genomics</p><p>In genomics, the ability to process and analyze large data sets is critical to advances in personalized medicine. TIG fosters collaboration among data scientists, biologists, and clinicians by providing a platform where computational methods for genome analysis can be developed and shared. For example, an innovator might develop an algorithm to identify genetic markers associated with specific diseases. This algorithm can be refined and tested by benchmarkers, leading to improved methods for early disease detection and personalized treatment plans.</p><p>Impact: TIG supports cross-disciplinary collaboration, enabling breakthroughs that come from the convergence of different fields. By creating a marketplace for computational methods, TIG facilitates the exchange of ideas and tools that drive innovation at the intersection of scientific disciplines.</p><p><strong>4. Increased efficiency in resource allocation</strong></p><p>Real-world example: high-energy Energy Physics</p><p>In high-energy physics, experiments often generate massive amounts of data that require sophisticated analysis methods. The Innovation Game can optimize resource allocation by enabling the development of more efficient algorithms for data processing. For example, innovators can develop methods for filtering and analyzing collision data from particle accelerators. Benchmarkers help identify the most effective methods through competition, which are then adopted and refined for use in actual experiments. This ensures that the most efficient and effective tools are used, maximizing the scientific return on investment.</p><p>Impact: By facilitating price discovery and value capture, TIG ensures that resources are directed to the most promising and effective computational methods. This leads to more efficient use of funds and time, ultimately accelerating scientific discovery.</p><p><strong>5. Supporting Open Science and Reducing Duplications</strong></p><p>Real-world example: Machine Learning in Medical Imaging</p><p>In medical imaging, machine learning algorithms are increasingly being used to aid in the diagnosis of disease. However, the proprietary nature of many algorithms leads to redundancy and inefficiency as different research groups independently develop similar methods. TIG’s open data license encourages the sharing of algorithms and data, reducing duplication of effort. Innovators who contribute to this open ecosystem can benefit from the collective improvements made by others, resulting in more robust and widely adopted tools.</p><p>Impact: TIG supports a culture of open science, reducing redundancy and encouraging the collective advancement of knowledge. This collaborative approach not only accelerates the development of new methods but also ensures that the benefits of innovation are shared across the scientific community.</p><h4><strong>Final Words</strong></h4><p>The Innovation Game represents a significant change in the way scientific research is conducted and computational methods are developed. By introducing a synthetic market for algorithms, facilitating price discovery, and capturing value through strategic licensing, TIG accelerates the pace of innovation, democratizes access to advanced tools, and fosters a collaborative environment that transcends disciplinary boundaries. TIG’s impact on science will be profound, leading to faster discoveries, more efficient research processes, and a more inclusive and collaborative global scientific community.</p><p><strong>For Further Information</strong></p><p>Website: <a href="https://www.tig.foundation/home">https://www.tig.foundation/home</a><br>X account: <a href="https://x.com/tigfoundation">https://x.com/tigfoundation</a><br>White Paper:<a href="https://www.tig.foundation/whitepaper">https://www.tig.foundation/whitepaper</a><br>Discord: <a href="https://discord.com/invite/dDGBXaHqDD">https://discord.com/invite/dDGBXaHqDD</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=482b8c9ed917" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Implications of Asymmetric Problems in Modern Science]]></title>
            <link>https://medium.com/@0x_Rorschach/implications-of-asymmetric-problems-in-modern-science-feb1c402e16c?source=rss-0b668e3fef6c------2</link>
            <guid isPermaLink="false">https://medium.com/p/feb1c402e16c</guid>
            <category><![CDATA[asymmetric]]></category>
            <category><![CDATA[problem-solving]]></category>
            <category><![CDATA[cryptocurrency]]></category>
            <category><![CDATA[innovation]]></category>
            <category><![CDATA[science]]></category>
            <dc:creator><![CDATA[Rorschach]]></dc:creator>
            <pubDate>Sun, 18 Aug 2024 07:44:38 GMT</pubDate>
            <atom:updated>2024-08-18T07:44:38.428Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*x8KUEoyt-NB6Lz-0iYI9eg.png" /></figure><p><strong>1. Introduction to Asymmetric Problems</strong></p><p>At first glance, discussions of “symmetry” and “asymmetry” seem completely out of place in a lecture on the nature of modern science or in a paper on contemporary scientific and technological revolutions. However, a deeper analysis may reveal that we have become accustomed to solving asymmetric problems using inappropriate (symmetric) methods. In this paper, I will refer to problems as “asymmetric” when their causes and goals are different. In other words, there are contexts in which a problem is satisfied if its goal is achieved (or comes closer to “real” satisfaction), but remains unsatisfied if its causes are eliminated or reduced to some extent. (1)<br>This makes it difficult to understand the complexity and dynamics of different systems and their interconnectedness. Few studies have focused on this aspect, although it seems that increased attention to this concept could have important implications in contemporary societies coping with rapid development and increased reliance on technological systems. The aim of the present study is to raise awareness of the importance of asymmetric problems in various discourses and practical contexts. In contrast to studies that focus on specialized definitions of asymmetric problems or theories, the more general goal of this study is to analyze the nature of the phenomenon itself without using narrow classifications for different disciplines. (2)</p><p><strong>1.1. Definition and Characteristics</strong><br>An asymmetric problem can be defined as a nonlinear, unequal, and non-symmetric distributional relationship between at least two relevant attributes or variables. Asymmetric problems can occur when distributions differ between aspects such as citizens in organized or unorganized (real or virtual) crime; regions, employment, and income; “normal,” commercial, or underground and organized economies; organizations and regulations and institutions; physical flows of goods, labor, capital, or services and regulations and institutions; or between different regulations themselves. Time, space, and spatiotemporal asymmetries are also possible. (3,4)</p><p>Asymmetric problems have several characteristics that allow the separation of self-organized systems into layers or differentiated groups of organized or unorganized criminals. The main characteristics of this type of asymmetric problem include the following: Depending on the chosen attributes (v), the relationship between them is unequal, non-symmetric and non-normal. The underlying distributions have a pronounced asymmetry, which may mimic to some extent a re-entry, a connected sequence, or a self-insurance system. The relationships being studied might change in ways that are not linear or normal, and they might change in ways that are not normal or linear in terms of space and time. This could lead to a dynamic, hierarchical distribution that can oscillate, like in period doubling scenarios (1:1, 1:2, 1:4). Temporal and spatial modulations, drifts, and trends may violate the constancy of relationships and thus their relevance and transferability in the space-time relationship. In addition, spatial differentiation may appear unbalanced. This may be the result of inequalities in the physical, natural, social, economic, and regulatory background, such as the availability of environmental, procedural, behavioral, or structural opportunities and constraints that support different types and levels of criminal organization. (5)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*2DF7RLIEyuAilfLM21VenQ.png" /></figure><p><strong>1.2. Examples in Different Scientific Disciplines</strong><br>Asymmetric problems can be found in virtually every corner of science and technology, requiring a wide variety of innovations. Such problems can be found in all branches of the natural sciences, such as chemistry, biology, physics, mathematics, computer science, geosciences, and advanced engineering disciplines, which may combine several of these. In the social sciences and humanities, they may arise in political ideology, civic and ethical norms, language and philosophy, communication and preference theory, choice under uncertainty, or evolutionary aspects. They may also occur in interdisciplinary contexts of user-driven research and practice at the interface of the human, social, and natural sciences, and may involve efforts to innovate new methodologies or new ways of applying developed methodologies to unsolved problems where there is an immediate need. Indeed, the proliferation of new “index” and “atlas” titles in recent years suggests that user-centered research and development activities in all kinds of agencies and organizations are addressing asymmetric problems at all scales. (6)<br>Problems that differ in such “asymmetric” ways have been addressed in many fields of science and technology. Asymmetric knowledge transfer is also the subject of much current research and discussion, especially in political theory and public administration. Asymmetry in computer science, mathematical relationships like “greater than,” computer science (like how hard it is to solve some problems compared to their positive or negative examples), operations research, problem complexity theory, causal decision theory, cut-off theorems, systems engineering and design, social systems, attribution theory, “determinacy” and constructive empiricism, cognitive neuroscience (like vision and executive attention), deductive modeling, developmental and learning programming, reasoning, and language. In statistical practice, stratified sampling is a type of asymmetric research. In fact, asymmetric problems and their solutions can be found in ordinary vital statistics applications, such as estimating disease frequencies and case-fatality rates, as well as in more complex analytical applications. (7)</p><p><strong>2. Challenges presented by asymmetric problems</strong><br>It is difficult to deal with asymmetric problems in science because conventional analytical methods are often inadequate for dealing with such problems. These challenges are multifaceted, involving uncertainty, behavioral change, and deep institutional and organizational issues. In addition, interdisciplinary work can be slow and productivity can be minimal, even futile. <br>Asymmetric problems-such as those in modern science, where principles are violated or new variables affect the validity of a particular set of laws or practices-are increasingly common. Traditional methods of methodological analysis may not be sufficient to deal with the methodological and technical developments that typically accompany asymmetric development, which can create more complicated external factors.<br>The overall complexity can lead to problems of collaboration and communication, as technical and practical terms may differ from context to context. As a result, scientists, innovators, and policymakers may not clearly see that they are addressing the same problem. To address these interdisciplinary challenges, theories of institutional change have invoked coordination as the elephant in the room. Often, studies use a variety of phrases, but need to bridge the gap between “could,” “should,” and “is” for change. The results may not even be clear or may be overtly close. (8)</p><p><strong>2.1. Methodological and Analytical Issues</strong><br>In a previous section, it was argued that the ways to create social welfare in a diverse world depend on how the different technological, economic, social, political, and cultural systems are treated. These systems can not be studied scientifically using the same assumptions and tools as physical laws. Because some system properties are peculiar and cannot be constructed under experimental control, they require different methodologies. This is indeed the case with asymmetric problems. Harries, ambiguities, and irreversibilities are quite important and are transferred from a socially meaningful level of events to consequences and the future. So, even if this position is sometimes blurred, as one might legitimately think that the world would be more organized and comfortable if most systems could behave like physical ones, we think that this particular kind of complexity needs to be treated in a different way. (9)</p><p>After the implementation of <strong>neo-Darwinian science in antiquity</strong> and the development of social dynamics in the modern world, very challenging fields of study became available. In many cases, however, biology (population genetics, immunology, signaling networks, ecosystem studies) and medicine (homeostatic regulation, immune responses, diagnosis of disabilities, syndromes, etc.) must also consider and deal with specific asymmetric complexities. Possible explanations for this mismatch may be related to the fact that asymmetric problems are much more difficult to study. When the true model of a process is described in asymmetric terms, it requires specific development of methodology and analytical mechanics. <br>Algorithms for numerical modeling and optimization also require specific designs, especially when the classical assumptions of uniqueness and smoothness of optima no longer hold. Thus, from different but related perspectives, many different and sometimes converging results lead to the same conclusions: there are too many models of complex science that make no sense, especially in this new direction of scientific modernity, and hardly any robust and emergent patterns of behavior are actually recognized. (10)</p><p><strong>2.2. Interdisciplinary cooperation and communication</strong><br>Perhaps the most important questions arising from the study of asymmetric issues concern the difficulties of interdisciplinary collaboration and communication. All available knowledge, no matter how seemingly irrelevant, can be useful to those working on smaller or symmetric problems; thus, in theory, they can be addressed by a single discipline. Similarly, the problems addressed by those working on larger or symmetric problems can be reduced to issues addressed by one or more disciplines. Asymmetric problems, however, are much less tractable because of the sheer variety of physical conditions, societies or social interactions, and the disparate, interacting aspects of the individuals involved. Given the highly specific details of individual cases, there is almost always a need for insights or information from many different fields of research. However, communication specialists are rarely in a position to produce or synthesize the necessary information because those working in different fields often have little common ground. (11)</p><p>Rather than face and work through this complexity, most individuals or collaborative groups rely on their own specializations and make assumptions about areas outside their considerable expertise. Political scientists, for example, very often assume the economic principles of certain forms of capitalism rather than doing their own economic research, consulting with those who do, or reviewing studies in the field. Thus, there are always questions about the unknowns in communication between individuals or groups, as well as about the reliability and representativeness of any one interpreter/translator’s research or even the selected research of any one interpreter/translator.</p><p>Thus, the only reliable way to address asymmetric problems is through an interdisciplinary research program that extracts patterns and principles from the entire global (or, more likely, a relatively large local) sample of problems and solutions. Everyone involved needs to talk to each other and examine their assumptions, methodologies, and other characteristics of their work. the lack of interdisciplinary cooperation and communication is one of the major obstacles to creating a comprehensive way to mitigate asymmetric problems. (12)</p><p><strong>3. Impact on research and innovation</strong><br>Asymmetric problems cause unjust suffering and harm in the form of less knowledge, more mistakes, and waste of societal resources. It is clear that science, and especially research practices, are affected. In addition, there are other areas of impact. In excess, such harms may become clinical, scientific, ethical, and political reasons for considering them unacceptable. Since the main task of ethical decision-making is to clarify whether such harms should be allowed or avoided, policies and resource allocation will be affected. (13)<br>Another important implication is for policy and funding. A decision-maker should give high priority to grant applications that aim to solve part or all of the asymmetry problem. In line with programmatic research and the Maori concept of collective problem formulation, the policy response must be broad in character to match the complexity of an asymmetric problem. While plugging the leak on the Titanic may seem like a good idea to those on board who are staying afloat, the scale of the problem means that simply staying afloat is not enough.</p><p>We need data to learn more about the effects of asymmetric information. Such evidence must inform decisions about whether to develop treatments to improve compliance and effective interventions, or to focus on eliminating the practices that damage clinical information. A non-medical clinical trial can help make such inferences in the presence of symmetric evidence. The data in the intelligence intervention trial produce evidence about the overall harm or benefit of choosing a course of active intervention based on some symptomatic excoriating information. Ethical problems are therefore potentially part of the physiological research evidence on nonclinical information asymmetry. (14)</p><p><strong>3.1. Ethical considerations</strong></p><p>Although the technical and methodological implications of asymmetric problems in modern science are significant, the ethical dimensions are among the most critical. All solutions to an asymmetric problem are bound to be asymmetric in some way, and merely looking for standard solutions to such problems risks solving a less asymmetric problem by creating another. Any research or action that attempts to maneuver around or across the asymmetries will therefore, appear to work in ways that one system or another does not. Indeed, it is the very subject of different articulations of the asymmetries of difference that would motivate and justify attention to different dimensions in the broader asymmetric quest.</p><p>The problems, however, must also respond to a host of other questions that go beyond simply informing research on implicative questions and solutions. Unfortunately, decisions to undertake research that addresses such implicative dilemmas cannot avoid being shaped by such value-laden orientations, if for no other reason than those mentioned above. In recent years, a number of ethical frameworks and guidelines have provided guidance in this area. It is probably worth spending some time reviewing some of these in more detail as we begin to wade into the same forest of such considerations ourselves. The idea of “ethical steering” offers a way of working with the problems of responding ethically to the implications of dealing with asymmetries in science and society without assuming that this is a straightforward matter of deciding on universal rules and the like. (15)</p><p><strong>3.2. Policy and funding implications</strong></p><p>Most areas of scientific research face problems of asymmetric structural observability, at least to some extent, and the ability to make panoramic or holistic assessments of fields is severely limited. Some forms of research funding, such as center grants or investigator-initiated transdisciplinary projects, are better suited to successfully address asymmetric problems and to exploit complementarities and synergies across pillars or functional areas to find cost-effective, solution-oriented support. Science indicators to inform policy are also likely to be resources that are useful for both the asymmetric and symmetric dimensions, or that link to impact indicators developed for senior management and good practice initiatives in research performing organizations (RPOs). However, there may be strategic advantages to pooling money across sectors in a more or less focused way if senior program managers believe that problems of asymmetric observability are not paramount, that progress is parametric, and that impact is slow to emerge even when there is agreement on a direction. (16)</p><p>The corollary of asymmetric problems is that a consistent, multi-tiered, but proportionate approach is needed to maximize the scientific returns and private and public benefits from research. It will be fruitful to continue to invest in the development of metrics and other indicator systems to capture the multiple costs and benefits of innovative activities that are relevant to decision making by organizations, businesses, and investors. Researchers, research support managers, and policy makers need to make strategic investment and disinvestment decisions, carefully exploiting the economically and strategically critical contributions that “itinerant problem solvers” can make to the entire process, from insight generation to innovation, delivery, and diffusion. (17)</p><p><strong>4. Strategies for dealing with asymmetric problems</strong></p><p>Decoupling and avoiding investment in unsuccessful R&amp;D are critical to success. When strong asymmetries in the problem are expected, there is a great advantage to investing in asymmetric information to address the main sources of uncertainty in the fastest and least costly way. As a result, a number of strategies have emerged to circumvent, or at least successfully manage, these problems or to reduce overall R&amp;D investment. As with new diseases, many of these problems are asymmetric along all the major dimensions (which will be mentioned), so that in any study based on the available data, the results remain inconclusive because “absence of evidence (of an association) is not evidence of absence (of an association)”. The strategy that permeates much scientific research cannot solve these issues. Here, I propose four broad possible approaches that can circumvent (i.e., partially or completely discard) these theoretical and practical problems. (18)</p><p>*First, one can build an adaptive research framework to address a few questions and then stop. So far, this makes room for clinical research.</p><p>*Second, one can try a “halfway” approach (between the null and alternative hypotheses). Suppose there is high uncertainty, and one can accept that the verdict requires different levels of evidence. In practice, one might accept that one study may be sufficient evidence for a level 2 claim, but two studies are required for level 1. Freriks and Slaats suggest a kind of four-case matrix that may help to categorize this and reconstruct the gradations of evidence.</p><p>*Third, as part of a data-sharing initiative, researchers worldwide can be invited to participate in a simple meta-analysis of the anonymized data.</p><p>*Fourth, to avoid fruitless discussions in the face of great uncertainty, methodologies/approaches could be developed to get around this problem. For example, a variety of statistical, machine learning, or artificial intelligence approaches can predict unknown phenomena when heterogeneity is also examined.</p><p><strong>4.1. Adaptive research frameworks</strong></p><p>There are research frameworks that are specifically designed to address asymmetric problems. The adaptive research framework has been used in several fields, including climate change science. This experience demonstrates the importance of being able to respond flexibly to increased levels of complexity or uncertainty when dealing with horizontal interactions. While it may be difficult to apply asymmetric research elsewhere, adapting research to address findings could increase the resilience of scientific pathways. Obviously, we can learn from research responses to serious diseases because they also deal with high complexity and uncertainty. The problem is not the existence of problems in scientific activities, but the ability of research to respond to, evaluate and cope with challenges. (19)</p><p>Adaptation is a key word in dealing with asymmetric problems; we need to be able to adapt. However, the assumption that we can adapt to the situation does not provide guidance in practice, for example, by making negative statements. Developing such requirements requires framing and, at the same time, an explicit decision about what the goals should be. Given the uncertainty and complexity of asymmetric problems, this means that these frameworks need to be adaptive in the sense that they should be flexible, because decisions and procedures are often controversial, and there are commitments about essential uncertainties and a wealth of interpretations. This further complicates the adaptive requirement. From the point of view of science, one could address these situations, as researchers attempt to decouple scientific knowledge and to develop the art of combining scientific and other knowledge with the powerful argument of the present. (20)</p><p><strong>4.2. Data Sharing and Open Science Initiatives</strong></p><p>With the growth of research traditions that address asymmetric problems, it seems clear that one of the obstacles to solving such problems is the lack of data or systematic investigation. Technologically advanced societies are blessed (cursed?) with vast amounts of data that could, in principle, be used to address these asymmetries. However, such societies are also strongly fixated on individualism, personal privacy, and concerns about vested interests. Nevertheless, there is currently an encouraging trend in science towards open access and open science. The modern emphasis on transparent methods, data sharing, and public access to knowledge is a reactive counter to this, fostering political, social, and systemic change in how science is done and what we can reasonably expect from it. (21)</p><p>Rather than embracing a principle of free inquiry by everyone for everyone, open science promotes an ethos of distributed collaboration as a means to an end. There are several advantages to such an approach. One is that by pooling knowledge and resources, we can circumvent a variety of different forms of asymmetric access. Because knowledge is shared among participants, we guard against the possibility that some researchers will make off with the results of others without giving sufficient credit. By publishing data and result sets in an organized and consistent framework (e.g., following a common standard such as BIDS or FCP), the sum of the data is far greater, and we are able to maintain a far more comprehensive window on what has been systematically studied and what has not. (22)</p><p><strong>5. Conclusion</strong></p><p>Asymmetric problems are ubiquitous in modern science and engineering. They pose barriers to applications and require the collaborative efforts of the community, including researchers, practitioners, and stakeholders in science and engineering, to address. This special issue of AI Innovation aims to present state-of-the-art research on understanding and tackling asymmetric problems in modern science. In this editorial, i first review the challenges and opportunities in dealing with asymmetric problems. Then, a summary of recent research is made to show the potential of dealing with asymmetric problems.</p><p>Finally, a brief overview of the papers in this special collection is given.Since sensor devices can only imperfectly capture complex real-world phenomena, noise in the sensed data is very common. As a result, problems with incomplete, biased, or misinterpreted observations are ubiquitous in various scientific domains, from healthcare to cybersecurity. This has led to a movement to adopt imperfect information from the base case, where uncertainty is assumed to be negligible. A concrete example is image recognition. Our conclusion is that asymmetric problems follow the call of the data-driven era. Therefore, in order to address asymmetric problems, it is imperative to stimulate communication between different scientific fields, each of which critically uses different data sources, processing tools, and system subcomponents and/or has different goals or requirements.</p><p><strong>References</strong></p><p>1. Arslan F, Singh B, Sharma DK, Regin R, Steffi R, Rajest SS. Optimization technique approach to resolve food sustainability problems. In2021 International Conference on Computational Intelligence and Knowledge Economy (ICCIKE) 2021 Mar 17 (pp. 25–30). IEEE. <a href="https://ieeexplore.ieee.org/abstract/document/9410735/">[HTML]</a></p><p>2. Kong XZ, Postema MC, Guadalupe T, de Kovel C, Boedhoe PS, Hoogman M, Mathias SR, Van Rooij D, Schijven D, Glahn DC, Medland SE. Mapping brain asymmetry in health and disease through the ENIGMA consortium. Human brain mapping. 2022 Jan;43(1):167–81. <a href="https://onlinelibrary.wiley.com/doi/pdf/10.1002/hbm.25033">wiley.com</a></p><p>3. Brown T, Elshaer A. Pounding of structures at proximity: A state-of-the-art review. Journal of Building Engineering. 2022. <a href="https://www.sciencedirect.com/science/article/pii/S2352710222000043">[HTML]</a></p><p>4. Porfirev A, Khonina S, Kuchmizhak A. Light–matter interaction empowered by orbital angular momentum: Control of matter at the micro-and nanoscale. Progress in Quantum Electronics. 2023. <a href="https://www.sciencedirect.com/science/article/pii/S0079672723000083">[HTML]</a></p><p>5. Mišić J, Kemiveš A, Ranđelović M, Ranđelović D. An Asymmetric Ensemble Method for Determining the Importance of Individual Factors of a Univariate Problem. Symmetry. 2023. <a href="https://www.mdpi.com/2073-8994/15/11/2050/pdf">mdpi.com</a></p><p>6. Rasoolimanesh SM, Ringle CM, Sarstedt M, Olya H. The combined use of symmetric and asymmetric approaches: partial least squares-structural equation modeling and fuzzy-set qualitative comparative analysis. International Journal of Contemporary Hospitality Management. 2021 Jul 6;33(5):1571–92. <a href="https://www.emerald.com/insight/content/doi/10.1108/IJCHM-10-2020-1164/full/html">[HTML]</a></p><p>7. Vosooghidizaji M, Taghipour A, Canel-Depitre B. Supply chain coordination under information asymmetry: a review. International Journal of Production Research. 2020 Mar 18;58(6):1805–34. <a href="https://www.academia.edu/download/94051442/00207543.2019.168570220221111-1-h7flgj.pdf">academia.edu</a></p><p>8. Seeber I, Waizenegger L, Seidel S, Morana S, Benbasat I, Lowry PB. Collaborating with technology-based autonomous agents: Issues and research opportunities. Internet Research. 2020 Feb 3;30(1):1–8. <a href="https://www.researchgate.net/profile/Stefan-Morana/publication/339485741_Collaborating_with_technology-based_autonomous_agents_Issues_and_research_opportunities/links/5ebd954692851c11a867bd74/Collaborating-with-technology-based-autonomous-agents-Issues-and-research-opportunities.pdf">researchgate.net</a></p><p>9. Zheng C, You SL. Advances in catalytic asymmetric dearomatization. ACS Central Science. 2021. <a href="https://pubs.acs.org/doi/pdf/10.1021/acscentsci.0c01651">acs.org</a></p><p>10. Han B, He XH, Liu YQ, He G, Peng C, Li JL. Asymmetric organocatalysis: an enabling technology for medicinal chemistry. Chemical Society Reviews. 2021;50(3):1522–86. <a href="https://pubs.rsc.org/en/content/articlehtml/2021/cs/d0cs00196a">[HTML]</a></p><p>11. Bender K. Research–Practice–Collaborations in International Sustainable Development and Knowledge Production: Reflections from a Political-Economic Perspective. The European Journal of Development Research. 2022. <a href="https://link.springer.com/content/pdf/10.1057/s41287-022-00549-7.pdf">springer.com</a></p><p>12. Marinho A, Couto J, Teixeira J. Relational contracting and its combination with the BIM methodology in mitigating asymmetric information problems in construction projects. Journal of Civil Engineering and Management. 2021 Apr 16;27(4):217–29. <a href="https://journals.vilniustech.lt/index.php/JCEM/article/download/14742/10500/">vilniustech.lt</a></p><p>13. Currie G, Rohren E. Social asymmetry, artificial intelligence and the medical imaging landscape. Seminars in Nuclear Medicine. 2022. <a href="https://www.sciencedirect.com/science/article/pii/S0001299821000982">[HTML]</a></p><p>14. Chater N, Loewenstein G. The i-frame and the s-frame: How focusing on individual-level solutions has led behavioral public policy astray. Behavioral and Brain Sciences. 2023. <a href="https://wrap.warwick.ac.uk/168582/1/WRAP-i-frame-s-frame-how-focusing-individual-level-solutions-led-behavioral-public-policy-astray-22.pdf">warwick.ac.uk</a></p><p>15. Boada JP, Maestre BR, Genís CT. The ethical issues of social assistive robotics: A critical literature review. Technology in Society. 2021. <a href="https://www.sciencedirect.com/science/article/pii/S0160791X21002013">sciencedirect.com</a></p><p>16. Alam Z, Sun L, Zhang C, Su Z, Samali B. Experimental and numerical investigation on the complex behaviour of the localised seismic response in a multi-storey plan-asymmetric structure. Structure and infrastructure engineering. 2021 Jan 2;17(1):86–102. <a href="https://www.tandfonline.com/doi/abs/10.1080/15732479.2020.1730914">[HTML]</a></p><p>17. Saravanan V, Berman GJ, Sober SJ. Application of the hierarchical bootstrap to multi-level data in neuroscience. Neurons, behavior, data analysis and theory. 2020;3(5). <a href="https://www.ncbi.nlm.nih.gov/pmc/articles/PMC7906290/">nih.gov</a></p><p>18. Phillips W, Roehrich JK, Kapletia D. Responding to information asymmetry in crisis situations: innovation in the time of the COVID-19 pandemic. Public Management Review. 2023. <a href="https://www.tandfonline.com/doi/pdf/10.1080/14719037.2021.1960737">tandfonline.com</a></p><p>19. Turner MG, Calder WJ, Cumming GS, Hughes TP, Jentsch A, LaDeau SL, Lenton TM, Shuman BN, Turetsky MR, Ratajczak Z, Williams JW. Climate change, ecosystems and abrupt change: science priorities. Philosophical Transactions of the Royal Society B. 2020 Mar 16;375(1794):20190105. <a href="https://royalsocietypublishing.org/doi/pdf/10.1098/rstb.2019.0105">royalsocietypublishing.org</a></p><p>20. Aben TA, van der Valk W, Roehrich JK, Selviaridis K. Managing information asymmetry in public–private relationships undergoing a digital transformation: the role of contractual and relational governance. International Journal of Operations &amp; Production Management. 2021 Sep 8;41(7):1145–91. <a href="https://www.emerald.com/insight/content/doi/10.1108/IJOPM-09-2020-0675/full/pdf">emerald.com</a></p><p>21. Verschaffel L, Schukajlow S, Star J, Van Dooren W. Word problems in mathematics education: A survey. Zdm. 2020. <a href="https://www.researchgate.net/profile/Goenuel-Erhan-2/publication/356879570_Tracing_students&#39;_mathematics_learning_loss_during_school_closures_in_teachers&#39;_self-reported_practices/links/64bec389c41fb852dd98c2d7/Tracing-students-mathematics-learning-loss-during-school-closures-in-teachers-self-reported-practices.pdf">researchgate.net</a></p><p>22. Aubin I, Cardou F, Boisvert‐Marsh L, Garnier E, Strukelj M, Munson AD. Managing data locally to answer questions globally: The role of collaborative science in ecology. Journal of Vegetation Science. 2020 May;31(3):509–17. <a href="https://onlinelibrary.wiley.com/doi/pdf/10.1111/jvs.12864">wiley.com</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=feb1c402e16c" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>