<?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 Ardavan Khalij on Medium]]></title>
        <description><![CDATA[Stories by Ardavan Khalij on Medium]]></description>
        <link>https://medium.com/@ardavankhalij?source=rss-9c9ce9a2da80------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*KiSzDM8nN3RULXkXj-QSSA@2x.jpeg</url>
            <title>Stories by Ardavan Khalij on Medium</title>
            <link>https://medium.com/@ardavankhalij?source=rss-9c9ce9a2da80------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 21:56:57 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@ardavankhalij/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[Racism in artificial intelligence. Is this possible?]]></title>
            <link>https://ardavankhalij.medium.com/racism-in-artificial-intelligence-is-this-possible-81629047e30?source=rss-9c9ce9a2da80------2</link>
            <guid isPermaLink="false">https://medium.com/p/81629047e30</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[ml-so-good]]></category>
            <category><![CDATA[ai]]></category>
            <dc:creator><![CDATA[Ardavan Khalij]]></dc:creator>
            <pubDate>Wed, 20 Jul 2022 09:10:54 GMT</pubDate>
            <atom:updated>2022-07-20T10:22:24.787Z</atom:updated>
            <content:encoded><![CDATA[<h3>Abstract</h3><p>Artificial intelligence (AI). Undoubtedly, you must have heard about it. Today, it is everywhere. We use it in everything without even noticing it. However, with AI becoming more popular, some problems appeared. One of these problems is racism, as Julia Bossmann and Brian Rieger raised in their paper in “Weforum” [1] and “Labelbox” [2]. Racism is not a new concept, and it still exists in the modern world. Unfortunately, AI algorithms can also become racist because of their environment, the people around them, and the bias of some events. This can cause danger and injustice for some people or even hazardous and even a fatal problems.</p><p>Keywords: Artificial intelligence, Rasict, Racism, Injustice</p><h3>1. Discussion</h3><p>How does AI works? As Stuart J. Russell and Peter Norvig mentioned in their book, “Artificial Intelligence, a modern approach” [3] , AI works based on data it receives from the environment. First, it gets data from the environment via sen- sors, cameras, and other information processing tools. Then, it decodes them to a language that computers are familiar with, 0 and 1. After that, it makes decisions based on these data and the old data it received and then improves itself based on the information it receives over time.</p><p>So how can AI become racist? There can be several ways, but two are more hazardous than the others. The first way is that, unfortunately, there are still some racist people in our society. As Ernst Mach said, “We must not forget that all things in the world are connected with one an- other and depend on one another”. These people can be in any position. One of these positions is a computer scientist. So if someone related to an AI project, be racist, it can affect the judg- ment of the AI. For example, suppose he or she makes some bias on a race in the AI algorithm.</p><p>In that case, this algorithm can clearly become a racist algorithm. But, can an AI algorithm become racist itself? The answer is yes, it can. Even if the algorithm operated flawlessly, assume a crime detection algorithm based on AI to help the police force find criminals. Two possibilities can make this algorithm behave like a racist al- gorithm. The first possibility is that there can be racism in society and the algorithm learn from it. For example, maybe a few racist police forces exist, so the algorithm can learn from them. So even with a good algorithm, there is a possibility that the algorithm becomes racist because of the racism in society. The second possibility is that the algorithm becomes racist by accident. For example, in the previous example, there can be an accidentally higher crime rate for a group of peo- ple with a color of skin or a color of hair or even a nationality in a year, so the algorithm makes a bias for that group, which is obviously racism. Is there a solution? As Jane E. Fountain wrote in her paper, “The moon, the ghetto and artificial intelligence: Reducing systemic racism in com- putational algorithms” [4] , there are some ways to reduce this problem and also solve it. How- ever, the best way to reduce it in any aspect is less racism in society. There are several examples of this problem in real life. One of them that is raised by Lel Jones in “A Philosophical Analysis of AI and Racism” [5] is “racism in law enforce- ment AI”. One of the AI systems available now is COMPOS used by the police of the U.S. It checks the possibility of doing the same crime by an ar- rested person in the future based on their history. This system is designed to be fairer, but some cases show us the opposite. So there are many problems ahead of us. Some may say that these systems make racism less. They are correct, but racism is still a big problem. This system could only make a bad situation a less bad situation.</p><h3>2. Conclusions</h3><p>Overall, AI can solve many difficulties, but there can be some problems, and racism is one. So a programmer, a computer scientist, or an AI expert should be careful about what program he or she is writing because with some minor changes in their program, they can reduce the problem and prevent some irreversible events.</p><h3>References</h3><p>[1] J. Bossmann, Top 9 ethical issues in artificial intelli- gence, weform.org, Accessed, 2016.<br>URL <a href="https://www.weforum.org/agenda/2016/10/">https://www.weforum.org/agenda/2016/10/</a> top-10-ethical-issues-in-artificial-intelligence/</p><p>[2] B. Rieger, Labelbox Speaks on Ethics of AI at O’Reilly’s Strata Data Conference, Labelbox. Avail- able (2019).</p><p>[3] S.J.Russell,P.Norvig,ArtificialInteligenceAmodern approach, Pearson, Saddle River, New Jersey, 2010.</p><p>[4] J. E. Fountain, The moon, the ghetto and artificial intelligence: Reducing systemic racism in computa- tional algorithms, Government Information Quarterly (2021).</p><p>[5] L. Jones, A Philosophical Analysis of AI and Racism, Stance: An International Undergraduate Philosophy Journal 13 (2020) 36–46.</p><p><a href="https://medium.com/mlearning-ai/mlearning-ai-submission-suggestions-b51e2b130bfb">Mlearning.ai Submission Suggestions</a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=81629047e30" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Fuzzy logic Part 2 — (β, Cf, and find the right rule for new data)]]></title>
            <link>https://ardavankhalij.medium.com/fuzzy-logic-part-2-%CE%B2-cf-and-find-the-right-rule-for-new-data-81a4dd6b2dc3?source=rss-9c9ce9a2da80------2</link>
            <guid isPermaLink="false">https://medium.com/p/81a4dd6b2dc3</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[fuzzy-logic]]></category>
            <dc:creator><![CDATA[Ardavan Khalij]]></dc:creator>
            <pubDate>Sat, 28 Aug 2021 14:11:34 GMT</pubDate>
            <atom:updated>2021-08-28T14:36:22.998Z</atom:updated>
            <content:encoded><![CDATA[<h3>Fuzzy logic Part 2 — (β, Cf, and find the right rule for new data)</h3><p>In the first part of this story, I explained Fuzzy logic and defined the μ parameter, an essential parameter in Fuzzy logic. In this part, we focus on β, Cf and find the correct rule for new data.</p><h3>β parameter</h3><p>From the beginning, the goal was to classify the data, so the available laws each specify a category.</p><p>For example, if a person is overweight and has high blood pressure and high blood fats, they are 97% more likely to have a heart attack.</p><p>Whether this rule is more appropriate for the “heart attack” answer or for the “no heart attack” answer is determined using the β parameter, which is calculated using the following formula:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/798/1*o0aXTBBrKWmLyaZs_UTBSQ.png" /></figure><p>Where h is one of the classes, and the rest is exactly like the formula for calculating μ in <a href="https://medium.com/mlearning-ai/fuzzy-logic-part-1-the-%CE%BC-parameter-ad2f647cdca8"><strong>Fuzzy logic Part 1— (μ parameter)</strong></a>. In the above formula, the set of degrees of compatibility of all data is measured by one of the rules to obtain the most accurate case for the rule.</p><p>Now, to obtain the most consistent class with the law, according to the following formula, the maximum value of β obtained for each class is received, and thus the class with the highest β is selected as the law class extracted. Be.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*njFZ9-DsW9VhYAACoV-Teg.png" /></figure><h3>Cf parameter</h3><p>This parameter is known as the degree of confidence, and its general meaning is to obtain the degree of certainty of each law and indicates that according to the data based on which the law is determined, the degree of certainty for the law is determined. This parameter is obtained through the following formula:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/742/1*6KmQnfOJIrXOqNEAs-fYIA.png" /></figure><p>Where the following formula calculates β ̅:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/646/1*Aa8Z0HcF-nhvIXDSc2TpPg.png" /></figure><p>Where c is the number of classes.</p><h3>Find the right rule for new data</h3><p>When new data is examined, and the goal is to find the law that must be followed, the following formula is used to find the maximum degree of belonging in the degree of certainty, so the law that has that maximum is appropriate for the new data. Will be.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*_ScIRzfFeJOSTCBx-9oY4g.png" /></figure><p>Where S is the set of rules.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=81a4dd6b2dc3" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Fuzzy logic Part 1 — (the μ parameter)]]></title>
            <link>https://ardavankhalij.medium.com/fuzzy-logic-part-1-the-%CE%BC-parameter-ad2f647cdca8?source=rss-9c9ce9a2da80------2</link>
            <guid isPermaLink="false">https://medium.com/p/ad2f647cdca8</guid>
            <category><![CDATA[fuzzy-logic]]></category>
            <category><![CDATA[μ]]></category>
            <dc:creator><![CDATA[Ardavan Khalij]]></dc:creator>
            <pubDate>Fri, 27 Aug 2021 06:42:44 GMT</pubDate>
            <atom:updated>2021-08-27T07:12:42.007Z</atom:updated>
            <content:encoded><![CDATA[<h3>Fuzzy logic Part 1 — (the μ parameter)</h3><p>In fuzzy mathematics, fuzzy logic is a type of logic in which the actual value of the variables may be any real number between 0 and 1. This is used to control the concept of partial truth, where the amount of truth may be completely true and completely false. In contrast, in Boolean logic, the actual values ​​of the variables may be just integers 0 or 1.</p><p>Fuzzy logic is based on the observation that people make decisions based on inaccurate and non-numerical information. Fuzzy models or sets are a mathematical tool for displaying ambiguity and inaccurate information. These models can identify, represent, manipulate, interpret and use incomplete and uncertain data and information.</p><p>A master program may specify different subsets of a continuous variable. For example, a temperature gauge for anti-lock brakes may have several separate membership functions that define the specific temperature ranges required to control the brakes properly. Each function plots the same temperature value with a real value in the range of 0 to 1. These real values ​​can then be used to determine how the brakes are controlled. Thus, the fuzzy set theory provides a tool for representing uncertainty.</p><p>In fuzzy logic, variables called linguistic variables are used, which is defined as follows:</p><p>While variables in mathematics usually take numerical values, in applications of fuzzy logic, non-numerical values ​​are often used to facilitate the expression of rules and facts. A linguistic variable such as age may accept values ​​such as young and contrasting old. Because natural languages ​​do not always have enough value terms to express a fuzzy value scale, it is common to modify language values ​​with adjectives or adverbs. Fuzzy operations can plot mathematical input values ​​in fuzzy membership functions.</p><p>The process of assigning a system numeric input to fuzzy sets with degrees of membership is called fuzzy. This membership may be anywhere in the range [0,1]. If it is 0, the value does not belong to the given fuzzy set, and if it is 1, the value is completely in the fuzzy set. Any value between 0 and 1 indicates the degree of uncertainty that the value belongs to the set. These fuzzy sets are usually described in words. For example, the figure below shows the meaning of the terms cold, hot, and hot with temperature scale mapping functions. A point on this scale has three “truth values” one for each of the three states. The vertical line in the figure below shows the specific temperature measured by the three arrows. Since the red arrow is zero, this temperature may be interpreted as “not hot”. That is, this temperature is zero in the “hot” fuzzy set. The orange arrow (referring to 0.2) may describe it as “slightly warm” and the blue arrow (indicating 0.8) as “relatively cold”. Therefore, this temperature has 0.2 members in the “hot” fuzzy set and 0.8 members in the “cold” fuzzy set. The membership rate set for each fuzzy set is the result of fuzzy.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ycUs9qTUeSxP0lT9-4CZbA.png" /><figcaption>Temperature fuzzy logic diagram</figcaption></figure><p>In fuzzy logic, several vital parameters are the basis of fuzzy logic and implementation classification algorithm based on them, which are introduced and explained in order, and finally, how to conclude according to the rules.</p><h3><strong>μ parameter</strong></h3><p>As you know, in the proposed algorithm, evolutionary algorithms and fuzzy logic are used to extract the rules. Factors that help in the process of extracting these rules are used to calculate and extract them.</p><p>The first parameter is the μ parameter, which indicates the degree to which a data belongs to a category because each rule contains several parameters. Each parameter with a membership percentage equal to 100 × μ belongs to the category in which the temperature example was given. Now, if the membership rate of a data with a specific rule is the goal, you need to use the following formula:</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*FYei5__TQfW3ryabSuXaYA.png" /><figcaption>The formula for calculate μ</figcaption></figure><p>In this formula, Rj represents the law of j and represents the various factors of the law, which in this formula is one to one, and XP also represents the data whose compatibility is measured, which includes the factor. The method of obtaining each of the data parameters, whether from shapes or multi-criteria functions. This function can have specific forms that are the most convenient and common of these weaknesses, triangular correlation coefficient and trapezoidal weakness according to the data that may be different.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=ad2f647cdca8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The story of me studying Masters abroad — Part 2 (English Requirement)]]></title>
            <link>https://ardavankhalij.medium.com/the-story-of-me-studying-masters-abroad-part-2-english-requirement-fd134a9927c8?source=rss-9c9ce9a2da80------2</link>
            <guid isPermaLink="false">https://medium.com/p/fd134a9927c8</guid>
            <category><![CDATA[study-abroad]]></category>
            <category><![CDATA[toefl]]></category>
            <category><![CDATA[ielts]]></category>
            <category><![CDATA[toefl-test-preparation]]></category>
            <category><![CDATA[ielts-preparation]]></category>
            <dc:creator><![CDATA[Ardavan Khalij]]></dc:creator>
            <pubDate>Wed, 25 Aug 2021 04:02:45 GMT</pubDate>
            <atom:updated>2021-08-25T04:02:45.506Z</atom:updated>
            <content:encoded><![CDATA[<h3>The story of me studying Masters abroad — Part 2 (English Requirement)</h3><p>In this story, I focus on the English language Certification in applying for a master’s. Two English certificates are famous and acceptable in applying for a university, and they are IELTS and TOEFL. These two certifications are acceptable in almost all universities, although I should mention that there are some exceptions. For instance, Caltech does not accept IELTS, and they only accept TOEFL. Therefore, although TOEFL and IELTS are all suitable for all universities, you better do the TOEFL for universities in the USA or Canada.</p><p>IELTS can be suitable for all of your purposes. I study for my IELTS exam for two months, and I achieved 6.5 in total. IELTS has two types. Academic and General and the Academic is what you need for applying. The range of IELTS that you need to achieve is different for different universities and different courses. Nevertheless, I can sure you that you can not apply anywhere with IELTS less than 6.</p><p>IELTS has four parts. Listening, Reading, Writing, and Speaking. Most of the universities focus on writing and speaking, but they are all critical. So the test has four parts that three of them (reading, writing, and Listening) are at the same time, and one of them (speaking) is in another appointment.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/640/1*VT_RTXt9h779H_3rlWLR6A@2x.jpeg" /></figure><p>The listening test takes about 30 minutes. There are 40 questions in this section that you can choose from: Multiple choice, Matching, Graph or diagram, Note completion, table completion, summary, or short answers. This part of the IELTS test assesses your ability to understand concepts and information, identify ideas, tendencies, and goals of English-speaking people, and your ability to speak a subject.</p><p>The second part of the exam is Reading. This part of the test lasts about 60 minutes and assesses your ability to grasp the meaning of a text, the concepts, and the author. The reading section of the IELTS test itself is divided into two parts. Academic Department and General Department. The academic section includes long texts from books, magazines, and newspapers. The general section includes short general texts on the topics we deal with daily.</p><p>The third part is writing. This part of the test takes about 60 minutes. Like the reading, the part is divided into two parts, academic and general. The academic writing test itself consists of two questions. In the first question, you are given a diagram, chart, or table, and you are asked to explain or summarize it. Alternatively, you may even be asked to describe an event. In the second question, the academic section also asks you to write an essay in response to a topic or problem presented to you. The general text section also includes two questions: The first question puts you in a position and asks you to write a letter or text to describe the situation. Furthermore, in the second part of a text, you should write your opinion on a specific topic or problem.</p><p>Furthermore, the last part is speaking. The speaking test is a two-part test that lasts between 11 and 14 minutes. Candidates are interviewed individually by an expert, and the interview is recorded. In the first part of this test, the interviewer will ask you simple questions about yourself to prepare you for the second part, reducing your fear a little! General questions such as talking about your own family, where you come from, your education, or your interests… In the second part, which lasts between 3 and 4 minutes (including 1 minute of thinking about your answer), the interviewer gives you a task card and asks you to talk about it for 1 to 2 minutes. My advice is that talk about it for at least two minutes. The third part lasts between 3 to 4 minutes, in which the interviewer asks you a series of questions on the same card.</p><p>Each of these parts has a mark of 9, and at the end, your final mark is calculated by averaging your marks in different parts. For instance, if you got 6,7,6.5,7, your mark will be 26.5/4 = 6.625 that is near 6.5, so your mark is 6.5.</p><p>TOEFL also has its format. TOEFL is a four-hour computer test with four sections: Reading, Listing, Speaking, and Writing. You use a fundamental skill for each section (so in the reading section, you have to read texts, and in the listening section, you have to listen to audio clips). In some questions, you also need to use different skills. All sections have a timeline, and you always have 20 minutes in the speaking section and 50 minutes in the writing section. However, the reading and listing sections are different due to empirical questions. The reading section lasts 60 to 80 minutes, while the listening lasts between 60 and 90 minutes. No clues tell you which questions are empirical and not, so do not spend too much time worrying about it — do your best to answer all the questions! Finally, in the middle of the test, you will have a 10-minute break between the Listing and Reading sections. This is a mandatory time, which means you have to accept it. You can take other short breaks yourself during the test, but you can not spend this 10 minutes on other work. You may be taking notes on all parts of the test. The test center manager gives you paper and pens to use. You will also have a headset with a microphone to listen to audio clips and record your speaking responses.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/512/1*s_xEO0_IVSwaFc4C44VZnQ@2x.jpeg" /></figure><p>The first part you encounter in the TOEFL is the reading part. As mentioned above, the length of time this section lasts can vary. So if you do not receive any experimental questions, the reading will take 60 minutes and contain 42–36 questions. If you receive empirical questions, it will take 80 minutes and will contain 48 to 56 questions. Your main task is to read a text and answer related questions. Any text you see in the reading section; is lengthy — usually about 700 words — and you will see three or four texts in total (if you get four texts, one of them will be an experiment). The texts deal with a variety of academic topics, from history and art to science and psychology. Each text is accompanied by 12 to 14 questions (or a «set of questions»). You are not given hard times in every text and question, so it is crucial to managing your time wisely.</p><p>The second part is listening. Like the reading section, the length of time the listening lasts depends on the number of questions and the type of questions in this section. This section can last from 60 minutes with 34 questions to 90 minutes and 51 questions. Listening tests; Tests your English listening skills. In total, you will listen to four to six lecture texts and two to three conversations. Each lecture text lasts three to five minutes and is accompanied by six questions.</p><p>In comparison, each part of the conversation lasts about three minutes and has five questions. You only hear each audio clip once. When you listen to an audio file, you usually see an image of the speaker(s) on the screen.</p><p>The third part is speaking. Now is the time to use the microphone you have been given! The TOEFL speaking section takes about 20 minutes. For this part, you have to do six things. The first two parts, the independent part, ask you to talk for 45 seconds about your comments on topics of interest. The next four episodes are composite, and you have to talk about texts and audio clips for 60 seconds.</p><p>Finally, we come to the last part of TOEFL and that is writing. You should write two articles (a combined article and a standalone article) within 50 minutes. You will have 20 minutes for the first article and 30 minutes for the second article.</p><p>Overall, these two exams have a lot in common, and I think you should choose one that you are more comfortable with. This is the main advice I can give to you.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=fd134a9927c8" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The story of me studying Masters abroad — Part 1 (How to become prepare)]]></title>
            <link>https://ardavankhalij.medium.com/the-story-of-me-studying-masters-abroad-part-1-how-to-become-prepare-b59f6f40ab?source=rss-9c9ce9a2da80------2</link>
            <guid isPermaLink="false">https://medium.com/p/b59f6f40ab</guid>
            <category><![CDATA[gre]]></category>
            <category><![CDATA[computer-science]]></category>
            <category><![CDATA[abroad-education]]></category>
            <category><![CDATA[ielts]]></category>
            <category><![CDATA[university]]></category>
            <dc:creator><![CDATA[Ardavan Khalij]]></dc:creator>
            <pubDate>Tue, 24 Aug 2021 15:46:37 GMT</pubDate>
            <atom:updated>2021-08-24T15:46:37.242Z</atom:updated>
            <content:encoded><![CDATA[<h3>The story of me studying Masters abroad — Part 1 (How to become prepare)</h3><p>I was recently accepted into the Master of Computer Science at the Vrije Universiteit of Brussels. But, of course, I had many problems and challenges in this journey, and in this series of stories, I will share everything you need to know to study abroad and help you overcome these challenges. So, let’s get started.</p><p>For this journey, you need to prepare some documents and lists for yourself. These data can be different for each university.</p><p>If you aim to start your master’s in a university in Canada or the United States of America, you should begin searching for professors that work in a field you like. Generally, if you want to apply for a university in these two countries, you must apply for a direct Ph.D. Unfortunately, the costs of living and education in the USA and Canada are prohibitive. However, if you are wealthy enough, you can think about applying for a master’s.</p><p>You should also prepare a list of requirements each university wants, such as what GRE scores they need and what Toefl/IELTS score they need. Finally, of course, I should mention that if your aim is just these two countries, you better focus on Toefl because Toefl is superior in Canada and the USA.</p><p>So applying for North America is more challenging than in Europe. For applying to Europe, you need to fill the application form and provide the documents each university needs. For funding in Europe, you should just be good enough. If you are good enough, they give you a scholarship or a fund. However, the cost of study in some countries in Europe, such as Switzerland, Belgium, Italy, and Germany, is very cheap.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/0*zxwqRMYgEe4r1OVL.png" /></figure><p>Meanwhile, it would be best if you prepared yourself for your GRE and Toefl/IELTS exams. Score 320 in GRE is perfect. If you get this score, you can be sure of getting a scholarship somewhere or having the support of a professor in the USA or Canada. Also, with overall IELTS Academic 6.5 or 7.0, you are safe for applying for almost anywhere. You also need to be careful about your GPA. The higher the GPA, the better your chances of being accepted.</p><p>You also need a standard CV, an SOP, your bachelor’s diploma, or a certificate that shows you are still studying for your bachelor’s and you will be finished until the date you want to start your master’s in the destination university, your transcripts of scores, and Recommendation letters from at least two of your professors. Of course, I should mention that having no Recommendation letter is better than a bad one.</p><p>The last think you should know is that you should apply for universities that are achivable to get in. this is an important issue becouse if you just apply everywhere, you should pay a lot of application fee.</p><p>Overall, You should start sending Emails to the Professors in Canada and the USA from October. This is the ideal time for doing that. However, you should also be careful with the deadlines, and don’t miss them. I also highly recommend that you prepare all the documents I mentioned before the end of October.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=b59f6f40ab" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[What is AI, Machine Learning, Data Mining, and Fuzzy Logic?]]></title>
            <link>https://ardavankhalij.medium.com/what-is-ai-machine-learning-data-mining-and-fuzzy-logic-8467dd9b5117?source=rss-9c9ce9a2da80------2</link>
            <guid isPermaLink="false">https://medium.com/p/8467dd9b5117</guid>
            <category><![CDATA[comparison]]></category>
            <category><![CDATA[data-mining]]></category>
            <category><![CDATA[fuzzy-logic]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Ardavan Khalij]]></dc:creator>
            <pubDate>Sat, 07 Aug 2021 14:50:50 GMT</pubDate>
            <atom:updated>2021-08-08T03:40:49.695Z</atom:updated>
            <content:encoded><![CDATA[<p>Today, data mining and artificial intelligence are widely used sciences. Today, our world is disrupted without them, from computer games to search engines and from social networks to online shopping stores. Fuzzy logic has long been an efficient tool. Today’s science requires fuzzy logic because it can be used to increase computational accuracy. The following is a general definition of all three.</p><p>The concept of extracting new information from existing information using almost reliable methods is called data mining. Data mining is a way to access new data and information that can be extremely useful. Data mining has several methods and types of algorithms.</p><p>Artificial intelligence refers to the intelligence shown by the machine, which is essentially algorithms that optimally learn and draw conclusions using them. Artificial intelligence uses the available data to model and learns and then draw conclusions using them. There are many types of artificial intelligence algorithms, one of which is an evolutionary algorithm.</p><p>Machine learning is a way to analyze data. Machine learning is a branch of artificial intelligence based on the idea that systems can learn from data, identify patterns, and make decisions with minimal human intervention.</p><p>Data mining is inextricably linked to artificial intelligence and machine learning, and without the two, it can not move forward. The relationship between artificial intelligence, machine learning, and data mining is such that if any of these areas. If it does not do its job properly, other areas do not provide the right results and patterns.</p><p>Data mining and machine learning, in addition to their similarities, also have differences. Machine learning aims to build a machine that can learn and can adapt to new information. Nevertheless, data mining is science for extracting knowledge from a data set. Data mining focuses on industrial issues and practical solutions that, in addition to concerns about data size, in terms of big or small data also focuses on data processing speed.</p><p>Fuzzy logic can be explained with an example. Humans use words like old and young, but what does it mean if we pay more attention to them? Is a person who is twenty years old young, but a person who is twenty years and one day old is considered old? Old age and youth do not change suddenly and are relative concepts. The meaning of fuzzy logic is that if a person is 20 years old, he is at the peak of youth, but the next day he does not suddenly grow old but gradually decreases his youth. So we can say that fuzzy logic, instead of considering the concepts as zeros and ones, considers them as real numbers between zeros and ones, which will increase the accuracy of the calculations.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=8467dd9b5117" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Evolutionary Computation — Part 2 (Different Sections and Abilities)]]></title>
            <link>https://ardavankhalij.medium.com/evolutionary-computation-part-2-different-sections-and-abilities-eaaaa3b60ef0?source=rss-9c9ce9a2da80------2</link>
            <guid isPermaLink="false">https://medium.com/p/eaaaa3b60ef0</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[evolutionary-computation]]></category>
            <category><![CDATA[data-science]]></category>
            <category><![CDATA[fuzzy-logic]]></category>
            <dc:creator><![CDATA[Ardavan Khalij]]></dc:creator>
            <pubDate>Sat, 10 Jul 2021 02:40:56 GMT</pubDate>
            <atom:updated>2021-07-16T16:32:08.252Z</atom:updated>
            <content:encoded><![CDATA[<h3>Evolutionary Computation — Part 2 (Different Sections and Abilities)</h3><p>Evolutionary Computation algorithms have seven steps:</p><p>1. Determine the structure of each member of the population</p><p>2. Primary population production</p><p>3. Evaluate the answers produced</p><p>4. Choosing parents</p><p>5. The production of children from selected parents</p><p>6. Selection of remainings</p><p>7. Repeat the algorithm until a termination condition is reached</p><p>In the first step, the structure of the population should become clear. For example, in the knapsack problem, the structure of the population members is a binary array with the size of the number of available objects. This structure could make the process of the algorithm significantly more straightforward in the following steps if it chose the right and proper.</p><p>The second step is to produce the primary population. This step should be done randomly, but the important thing is that the number of population members is not random, and it can be chosen.</p><p>In the third step, members of the population should be evaluated, so each member gets a number that is the value or fitness of that member. We need to choose the evaluation function right. Otherwise, we receive an optimum based on some other quality that we do not need.</p><p>After that, in the fourth step, parents should be chosen based on their values. There are various methods to do this. One of them that is the simplest is the Roulette wheel. It is based on the value of the members. For instance, The chance to choose a member with the value of two is twice the chance to choose a member with the value of one.</p><p>In the fifth step, the children should be produced. For producing children, there are two operators in the Evolutionary Computation algorithms. They are Crossover and Mutation. Each of these operators has different types. Crossover is combining two parents and creating new members. This operator is based on what happens when any person or animal is born. Our DNA is based on our parents’ DNAs. Therefore we have some qualities from each of our parents. The other operator is Mutation. After creating children, some of them can be chosen randomly for some changes in their qualities. For example, in the knapsack problem, one of the zeros of a new member can become a one, or the opposite can happen. Notice that crossover is the first operator that should be used in this step, and after creating all the next generation, Mutation can be used. This can help the functionality of Evolutionary Computation algorithms.</p><p>The sixth step is to choose the next generation from the population. In any population, some of the members will die at a point. In these algorithms as well, some of the members will be deleted from the population. In order to make the next generation better, members have a chance based on their value. For instance, the chance to choose a member with the value of two is half the chance to select a member with the value of one; therefore, more valuable members will remain for the next generation.</p><p>And finally, in the last step, we should find a good termination condition. This condition is a way to stop these algorithms. Most of the time, choosing the number of generations for this condition is a good choice for these algorithms; however, there are many choices.</p><p>Evolutionary Computation algorithms can be used in many different subjects. I believe that using Evolutionary Computation algorithms can be very useful in Data Mining. Rules can be extracted from a dataset with them and a little help from Fuzzy systems. I recently started to study this kind of algorithm, and I found them very interesting. I also started to study some of these algorithms, such as the Michigan approach, and I think they can become more beneficial by being a little simpler. One of the things that can become simpler in this algorithm is the Evaluation function. So I started to write a program to see if it is possible to make the Evaluation function simpler, and I got some acceptable results. My goal was to write a classification algorithm based on the Michigan approach. So after I wrote this program, I started to compare this algorithm with some of the classification algorithms in the SKL library in Python, and the accuracy was better in my newly developed algorithm.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/700/1*WO2BVITTUTNXDgKpd2vMYQ.png" /></figure><p>Overall, Evolutionary Computation is a fantastic kind of algorithm, with the advantages far outsizing the disadvantages. It can also give better results if it is combined with Fuzzy systems. I think Evolutionary Computation is the most useful subject I learned in my bachelor studies. It can be used quite successfully across many applications where other algorithms find a hard time. It is also a gleam that shows me what artificial intelligence can do. Interestingly, big things can be done even with a small part of AI.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=eaaaa3b60ef0" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Evolutionary Computation — Part 1 (Introduction and Benefits and weaknesses)]]></title>
            <link>https://ardavankhalij.medium.com/evolutionary-computation-part-1-introduction-and-benefits-and-weaknesses-d2212e2d0944?source=rss-9c9ce9a2da80------2</link>
            <guid isPermaLink="false">https://medium.com/p/d2212e2d0944</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[evolutionary-computation]]></category>
            <category><![CDATA[computer-science]]></category>
            <category><![CDATA[machine-learning]]></category>
            <dc:creator><![CDATA[Ardavan Khalij]]></dc:creator>
            <pubDate>Wed, 07 Jul 2021 07:35:09 GMT</pubDate>
            <atom:updated>2021-07-07T07:35:09.295Z</atom:updated>
            <content:encoded><![CDATA[<h3>Evolutionary Computation — Part 1 (Introduction and Benefits and weaknesses)</h3><p>In my view, Artificial Intelligence is one of the most practical fields of computer science. It is and will continue to be used across various applications such as data mining, IoT, and autonomous driving. It is fascinating how a computer can think and decide intelligently based on some algorithms that are not that complicated.</p><p>Artificial Intelligence can be used by numerous programming languages such as Swift, C++, Java, Prolog, Lisp, Pascal, Python, and almost every other programming language. I tried each of these programming languages and studied their structure and the way they work and compile or interpret. I find Python more suitable for AI because of the variety of its libraries; otherwise, I think scripting programming languages are not that safe for programming because making a mistake and not noticing it is much easier in this kind of programming language than a programming language like Pascal. Of course, this essay is about AI, so we move on to the main topic.</p><p>Computational Intelligence is of particular interest to me as one of the branches of AI. Computational Intelligence has three branches, namely Evolutionary Computation, Fuzzy Systems, and Neural Networks. Amongst these, I find Evolutionary Computation to be the most fascinating.</p><p>Evolutionary Computation is a family of algorithms for finding the global optimum. These algorithms are inspired by biological evolution. They are based on a random population, and they work with trial and error. The general idea of these algorithms is that when there is a population available, each population member has some qualities, and these qualities can be good or bad. If one of the population members has more bad qualities than others, the chance for survival becomes less than others for that member, and the possibility of making the next generation will be less than the others for that member. With this concept, the population will become better and better each generation because a better part of the generation will make the next generation.</p><p>Of course, Evolutionary Computation algorithms have both advantages and disadvantages. Their most important benefit is that they do not need much time to find the goal. The time complexity in this algorithm is not exponential, and it can be beneficial for solving many problems. For instance, solving the classic “knapsack problem,” an exponential time complexity problem, is much faster with these kinds of algorithms that do not have exponential time complexity.</p><p>Another critical advantage of Evolutionary Computation algorithms is that they have a reasonable space complexity. They only store a small amount of data at the beginning of the algorithm, and they do not add any other data after that. They only change the data in the following stages of the algorithm.</p><p>On the other hand, their main disadvantage is that they do not always give us the best answer, and they can be stuck in the local optimum. Of course, this can be reduced by changing some factors in the algorithm. In my opinion, this is a manageable disadvantage. Consequently, I think Evolutionary Computation algorithms are superb.</p><p>To be countinue…</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=d2212e2d0944" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Compare a New method of Classification with Methods that already exist.]]></title>
            <link>https://ardavankhalij.medium.com/compare-a-new-method-of-classification-with-methods-that-already-exist-3109f4f27150?source=rss-9c9ce9a2da80------2</link>
            <guid isPermaLink="false">https://medium.com/p/3109f4f27150</guid>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[fuzzy-logic]]></category>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[genetic-algorithm]]></category>
            <category><![CDATA[data-science]]></category>
            <dc:creator><![CDATA[Ardavan Khalij]]></dc:creator>
            <pubDate>Tue, 06 Jul 2021 03:22:52 GMT</pubDate>
            <atom:updated>2021-07-06T03:22:52.451Z</atom:updated>
            <content:encoded><![CDATA[<p>My Bachelor’s Thesis at the University of Kashan, under the supervision of Dr. Vahidipour, is about a new algorithm for classification and prediction. This project mainly compares the ability and accuracy of different classification methods. It compares them with a method developed with Genetic algorithms and fuzzy logic. We use python in this project and also some of its frameworks. for instense, pandas, matplotlib and sklearn. These frameworks are some of the best frameworks in this field. I hope you find this project useful.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/637/1*fxCzp-xWCzh_WfKWS3I4Ng.jpeg" /></figure><p>I should also say thank you to Dr.Negin Masoudi Alavi and also Kashan University of Medical Sciences for giving us these extremely valuable datasets to work with.</p><h3>We also use these methods</h3><h3>1. Logistic Regression:</h3><p>Definition: Logistic regression is a machine learning algorithm for classification. In this algorithm, the probabilities describing the possible outcomes of a single trial are modelled using a logistic function.</p><p>Advantages: Logistic regression is designed for this purpose (classification), and is most useful for understanding the influence of several independent variables on a single outcome variable.</p><p>Disadvantages: Works only when the predicted variable is binary, assumes all predictors are independent of each other and assumes data is free of missing values.</p><h3>2. Gaussian Naive Bayes:</h3><p>Definition: Naive Bayes algorithm based on Bayes’ theorem with the assumption of independence between every pair of features. Naive Bayes classifiers work well in many real-world situations such as document classification and spam filtering.</p><p>Advantages: This algorithm requires a small amount of training data to estimate the necessary parameters. Naive Bayes classifiers are extremely fast compared to more sophisticated methods.</p><p>Disadvantages: Naive Bayes is is known to be a bad estimator.</p><h3>3. Stochastic Gradient Descent:</h3><p>Definition: Stochastic gradient descent is a simple and very efficient approach to fit linear models. It is particularly useful when the number of samples is very large. It supports different loss functions and penalties for classification.</p><p>Advantages: Efficiency and ease of implementation.</p><p>Disadvantages: Requires a number of hyper-parameters and it is sensitive to feature scaling.</p><h3>4. K-Nearest Neighbours:</h3><p>Definition: Neighbours based classification is a type of lazy learning as it does not attempt to construct a general internal model, but simply stores instances of the training data. Classification is computed from a simple majority vote of the k nearest neighbours of each point.</p><p>Advantages: This algorithm is simple to implement, robust to noisy training data, and effective if training data is large.</p><p>Disadvantages: Need to determine the value of K and the computation cost is high as it needs to compute the distance of each instance to all the training samples.</p><h3>5. Decision Tree Classifier:</h3><p>Definition: Given a data of attributes together with its classes, a decision tree produces a sequence of rules that can be used to classify the data.</p><p>Advantages: Decision Tree is simple to understand and visualise, requires little data preparation, and can handle both numerical and categorical data.</p><p>Disadvantages: Decision tree can create complex trees that do not generalise well, and decision trees can be unstable because small variations in the data might result in a completely different tree being generated.</p><h3>6. Random Forest Classifier:</h3><p>Definition: Random forest classifier is a meta-estimator that fits a number of decision trees on various sub-samples of datasets and uses average to improve the predictive accuracy of the model and controls over-fitting. The sub-sample size is always the same as the original input sample size but the samples are drawn with replacement.</p><p>Advantages: Reduction in over-fitting and random forest classifier is more accurate than decision trees in most cases.</p><p>Disadvantages: Slow real time prediction, difficult to implement, and complex algorithm.</p><h3>7. Support Vector Machine:</h3><p>Definition: Support vector machine is a representation of the training data as points in space separated into categories by a clear gap that is as wide as possible. New examples are then mapped into that same space and predicted to belong to a category based on which side of the gap they fall.</p><p>Advantages: Effective in high dimensional spaces and uses a subset of training points in the decision function so it is also memory efficient.</p><p>Disadvantages: The algorithm does not directly provide probability estimates, these are calculated using an expensive five-fold cross-validation.</p><p>For more information, please click on this link:<br><a href="https://analyticsindiamag.com/7-types-classification-algorithms/">https://analyticsindiamag.com/7-types-classification-algorithms/</a> <br>I should also mention that I pick the explanation of each method from this website.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3109f4f27150" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[How to solve poor internet connection]]></title>
            <link>https://ardavankhalij.medium.com/how-to-solve-poor-internet-connection-99b0ef3f04e2?source=rss-9c9ce9a2da80------2</link>
            <guid isPermaLink="false">https://medium.com/p/99b0ef3f04e2</guid>
            <category><![CDATA[internet]]></category>
            <category><![CDATA[poor-connection]]></category>
            <category><![CDATA[webp]]></category>
            <dc:creator><![CDATA[Ardavan Khalij]]></dc:creator>
            <pubDate>Wed, 09 Jun 2021 14:33:17 GMT</pubDate>
            <atom:updated>2021-07-09T00:49:42.343Z</atom:updated>
            <content:encoded><![CDATA[<p>Poor internet connection is a big problem in some places in the world. There are many places in the world with this problem. There are some ways to make this problem a little easier to handle.</p><h3>1. Using WebP to reduce the image file size</h3><p>These days, most applications and websites have high graphical environments, and some have lots of high-quality images. Of course, high-quality images enhance the user experience and reduce the speed of application, and it can make the delay of loading longer. WebP can reduce the image file size between 30 to 70 percent, but it’s not noticeable with the naked eye.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/939/1*Vg1A6ruiEys8trVIs_ABkw.png" /></figure><h3>2. Asynchronous calls</h3><p>In this chapter of this paper, we are going to mention Asynchronous calls. We can queue and save all of the requests in this queue in a file when the internet is unavailable.<br>We can do this with python, but I do not think that writing the code here should be necessary.</p><h3>3. Update with intention</h3><p>One of the main points that we must know in application design is to Update with intention. This means that we should update our application only when necessary and when it can help make the abilities better. Not for times that we think something maybe looks better or more beautiful.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=99b0ef3f04e2" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>