<?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 Gevin Hasmitha Karunarathne on Medium]]></title>
        <description><![CDATA[Stories by Gevin Hasmitha Karunarathne on Medium]]></description>
        <link>https://medium.com/@gevinhasmitha?source=rss-9c1d718612ff------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/0*6vKUbsgOmR4HDAmv</url>
            <title>Stories by Gevin Hasmitha Karunarathne on Medium</title>
            <link>https://medium.com/@gevinhasmitha?source=rss-9c1d718612ff------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Wed, 18 Nov 2026 20:03:30 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@gevinhasmitha/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[Transformers Still Matter: Why the Foundation Still Stands Strong Amidst the LLM Surge]]></title>
            <link>https://medium.com/@gevinhasmitha/transformers-still-matter-why-the-foundation-still-stands-strong-amidst-the-llm-surge-66e7d9b439e6?source=rss-9c1d718612ff------2</link>
            <guid isPermaLink="false">https://medium.com/p/66e7d9b439e6</guid>
            <dc:creator><![CDATA[Gevin Hasmitha Karunarathne]]></dc:creator>
            <pubDate>Thu, 15 May 2025 13:07:58 GMT</pubDate>
            <atom:updated>2025-05-16T06:03:45.432Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*0_XPEfuHSyaIhzV0EYuONQ.png" /></figure><p>Over the last years, the world of AI has been captivated by the meteoric rise of Large Language Models (LLMs) like GPT-4, Gemini, Claude, and Mistral. These models have shattered previous benchmarks in natural language understanding and generation, sparking discussions about Artificial General Intelligence (AGI) and revolutionizing industries. However, beneath all the buzz, one thing remains surprisingly stable: the core technology powering these marvels is still the Transformer.</p><p>Despite all the innovations and variations being built atop it, the Transformer model — introduced in the landmark 2017 paper “Attention Is All You Need” — has not been dethroned. Here’s why Transformers still matter.</p><h3>1. Transformers Are the Foundation of LLMs</h3><p>Almost every major LLM in use today is built on the Transformer architecture. Whether it’s OpenAI’s GPT series, Google’s PaLM and Gemini, Meta’s LLaMA, or Anthropic’s Claude, the underlying mechanics of self-attention, positional encoding, and layer normalization remain central. While hyperparameters, training datasets, and optimization strategies have evolved, the Transformer block structure continues to serve as the backbone.</p><p>In short, LLMs are not replacements for Transformers — they are extensions and scale-ups of them.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*loesS_zYIcwWlVGRxf24MQ.png" /></figure><h3>2. The Modular Power of Attention</h3><p>The self-attention mechanism introduced by Transformers remains one of the most versatile innovations in AI. It allows models to dynamically focus on different parts of the input, giving rise to rich contextual understanding. No other architecture has matched the flexibility and scalability of attention at this level of performance.</p><p>Even newer research exploring memory-augmented models, retrieval-augmented generation (RAG), or agentic systems frequently incorporate attention in some form or build directly upon Transformer-based LLMs.</p><h3>3. Transformers Power More Than Just Text</h3><p>Though Transformers gained fame in NLP, their influence has spread across domains:</p><ul><li>In vision: Vision Transformers (ViTs) rival convolutional neural networks.</li><li>In biology: AlphaFold uses Transformer-like models to predict protein structures.</li><li>In reinforcement learning and robotics: Transformers are used to model trajectories and policies.</li></ul><p>This architectural versatility is a major reason why the Transformer remains a central figure in AI research.</p><h3>4. Simpler, Smaller Transformers Are Still Hugely Useful</h3><p>While LLMs are often hundreds of billions of parameters, smaller Transformer models still dominate in edge deployment, enterprise applications, and research. Models like DistilBERT, TinyBERT, and MobileBERT (all Transformer variants) are optimized for speed and efficiency while retaining strong performance for many real-world tasks like sentiment analysis, document classification, and question answering.</p><p>In other words, we don’t always need an LLM. For many use cases, a smaller Transformer does the job just fine — with lower costs and higher interpretability.</p><h3>5. Continued Research and Refinement</h3><p>The Transformer isn’t stagnant. Research is actively improving its efficiency and capabilities:</p><ul><li>FlashAttention accelerates training and inference.</li><li>Sparse attention and linear attention methods scale Transformers to longer sequences.</li><li>Hybrid models combine CNNs and Transformers to capture both local and global features efficiently.</li></ul><p>Rather than replacing Transformers, many of these improvements deepen their utility and make them more applicable to a wider range of problems.</p><h3>Conclusion: Don’t Count Out the Classics</h3><p>While we may be entering the age of agents and ever-larger LLMs, it’s important to recognize that these advancements stand on the shoulders of the Transformer. As research continues to refine and repurpose this powerful architecture, its presence will remain central to the evolution of AI systems.</p><p>The Transformer isn’t just still relevant — it’s still foundational.</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=66e7d9b439e6" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[Object Oriented Programming]]></title>
            <link>https://medium.com/@gevinhasmitha/object-oriented-programming-3927f5d2c46f?source=rss-9c1d718612ff------2</link>
            <guid isPermaLink="false">https://medium.com/p/3927f5d2c46f</guid>
            <category><![CDATA[oop-concepts]]></category>
            <category><![CDATA[series]]></category>
            <category><![CDATA[oop]]></category>
            <category><![CDATA[java]]></category>
            <dc:creator><![CDATA[Gevin Hasmitha Karunarathne]]></dc:creator>
            <pubDate>Sun, 03 Jul 2022 13:06:04 GMT</pubDate>
            <atom:updated>2022-07-03T13:26:14.044Z</atom:updated>
            <content:encoded><![CDATA[<h4>Part I — Objects and Classes</h4><p>This series of blogs will cover all the basics and implementation of the OOP concepts.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*xFt8c_BhKqk1XTp_Vhe_og.png" /></figure><h3>What is OOP?</h3><p>OOP is a set of concepts which promotes security, code-reusability, code maintenance, data redundancy and ect… making use of objects and classes.</p><h3>Objects</h3><p>An object is a self contained component which contains attributes and behaviors (also known as methods or functions). So what are attributes and behaviors? Lets take a dog. The dog will have a name, an age…, these (eg; Rex and 03 yrs) can be considered the attributes of the dog. The behaviors of the dog is that the dog can walk, it can bark, it can eat and so on. So in that sense the dog can be considered an object as the attributes and behaviors are contained within a single <em>entity</em>.</p><h3>Classes</h3><p>Classes can essentially be considered as the blueprint for creating objects. It defines the variables and methods common to all objects of a certain kind.</p><p>For example, take a bicycle, there may be thousands of other bicycles with different colors, but all of the thousands of bicycles are made from the same blueprint. So we can say that the blueprint is the class and a bicycle is an object, and you can create many bicycles with the same blueprint.</p><p>Example of a class;</p><pre>public class Person{</pre><pre>  //Private variable<br>  private String fullName;</pre><pre>//Constructor<br>  public Person(String name ){<br>      this.fullName = name;<br>  }<br>  //A basic method<br>  public String addNames( String name1, String name2 ) {<br>     return name1 + name2 ;<br>  }<br>  //getter and setter methods<br>  public String getFullName() {<br>     return fullName;<br>  }<br>  public void setFullname (String fullName) {<br>     this.fullName = fullName;<br>  }<br>}</pre><h4>Attributes</h4><p>Attributes in a class can be simply the data members that represent the different features of the class, they can be closely related with variables.</p><p><strong>Instance attribute</strong> -It is an attribute declared within the constructor in the class, are automatically instantiated when an object is created from a class and has a scope of within that object.</p><p>Below, “name” is an instance attribute.</p><pre>public Person( String name ){}<br>      </pre><p><strong>Class Attributes</strong> -The attributes defined in a class that can be shared or accessed by it’s objects and is defined outside the constructor.</p><p>The following can be considered a class Attribute.</p><pre>private String fullName;</pre><h4>Methods</h4><p>Methods are used by classes in order to modify or access it’s attributes (getters and setters) or are used to perform a specific operation. A method is the logic or procedure defined in the class to do something. For example a bicycle can brake, the internal mechanics or procedure of the brake is the method which will give the desired output which is stopping the bicycle. We will be diving deeper into these parts in the upcoming blogs.</p><pre>  //A basic method<br>  public String addNames( String name1, String name2 ) {<br>     return name1 + name2 ;<br>  }<br>  //getter and setter methods<br>  public String getFullName() {<br>     return fullName;<br>  }<br>  public void setFullname (String fullName) {<br>     this.fullName = fullName;<br>  }</pre><h4>Constructors</h4><p>Constructors are<strong> </strong><em>special methods (without return type)</em> which allow you to fully initialize the object state before it can be used. So constructors are used when creating new objects from the class.</p><p>To create a constructor, you should create a method with the same name as the class and without any return type. Variables that are needed to be initialized are placed inside the constructor.</p><pre>  //Constructor<br>  public Person(String name ){<br>      this.fullName = name;<br>  }</pre><p>The ‘this’ keyword means that the variable is referring it’s current instance.</p><h3>Creating Objects from a class</h3><p>We now know that objects are created using a class as their blueprint. Objects are created by calling the constructor which creates a new instance of the class. This instance can be considered an object.</p><p>Lets create a new object from the Person class we saw before.</p><pre>Person p = new Person(&quot;George&quot;);<br>Person q = new Person(&quot;Shawn&quot;);</pre><p>The ‘new’ keyword creates a new instance of the class ‘Person’ with the ‘fullName’ variable initialized to “George.” So now as ‘p’ is basically an object of the class Person it has all the attributes and methods of the class ‘Person.’ The same applies to object ‘q.’</p><p>So you can create any number of Person objects with different names. This is just a simple way you could use objects. Say you want to store the age, name, gender and nationality for 10 students. By using OOP you can create a single class containing the attributes of age, name, gender and nationality and create 10 objects each with their own age, name, gender and nationality using a loop. This is much more efficient than creating 10 classes from scratch.</p><p>We will be discussing these in more depth in the upcoming blogs.</p><p><em>Please feel free to point out any mistakes that I’ve made.</em></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=3927f5d2c46f" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>