<?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 Ganesh Bharathwaj on Medium]]></title>
        <description><![CDATA[Stories by Ganesh Bharathwaj on Medium]]></description>
        <link>https://medium.com/@msdganesh?source=rss-2aee1599f169------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*Up7ULMJ4ZNOqskLPxZBNFA.jpeg</url>
            <title>Stories by Ganesh Bharathwaj on Medium</title>
            <link>https://medium.com/@msdganesh?source=rss-2aee1599f169------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Tue, 17 Nov 2026 02:40:04 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@msdganesh/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[Barcelona data engineer project]]></title>
            <link>https://msdganesh.medium.com/barcelona-data-engineer-project-600e20a21de2?source=rss-2aee1599f169------2</link>
            <guid isPermaLink="false">https://medium.com/p/600e20a21de2</guid>
            <category><![CDATA[data-engineering]]></category>
            <category><![CDATA[tableau]]></category>
            <category><![CDATA[la-liga]]></category>
            <category><![CDATA[etl]]></category>
            <category><![CDATA[fc-barcelona]]></category>
            <dc:creator><![CDATA[Ganesh Bharathwaj]]></dc:creator>
            <pubDate>Thu, 30 Mar 2023 09:02:38 GMT</pubDate>
            <atom:updated>2023-05-11T09:38:58.257Z</atom:updated>
            <content:encoded><![CDATA[<h3>FC Barcelona data engineering project</h3><p>I have developed a DE project which collects FC Barcelona’s(FCB) stats , transforms and loads it in a excel file. These data(stats) are displayed as chats in Tableau dashboard. I have upload my project in github (<a href="https://github.com/ganeshbhrathwaj/Barca_ETL">https://github.com/ganeshbhrathwaj/Barca_ETL</a>)</p><figure><img alt="My github repo" src="https://cdn-images-1.medium.com/max/913/1*m5-9dA908c9S_c-scsb6cw.png" /></figure><p>Extract.py :- This script will extract the stats of FCB from official Laliga website using selenium web-scraping. After extracting, it will load the data in a excel file (barca_stats.xlsx)</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/739/1*p7W0RHoxpxRcC_r0thp8gA.png" /><figcaption>Extract.py</figcaption></figure><p>Trasform.py :- This script is used to modify the data by removing the spanish accent letters to English letters. For example, it will convert names like <strong>OUSMANE DEMBÉLÉ to OUSMANE DEMBELE</strong></p><figure><img alt="" src="https://cdn-images-1.medium.com/max/832/1*jqLJIaC7tHJDj0aM7KfQXA.png" /><figcaption>Transform.py</figcaption></figure><p>ETL.py :- This performs the ETL by importing Extract.py and Transform.py methods one after the other</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/750/1*1hHptZ_8EiJ2T3AJAb8BHA.png" /><figcaption>ETL.py</figcaption></figure><p>I have published my Tableau workbook where the chats are displayed which uses barca_stats.xlsx as data source. This workbook has interactive dashboards.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*yghSv95zJ14O_ZWkkKp9iQ.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*55WWGiGa-G8DSRqAOIaGAQ.png" /></figure><p>You can visit my tableau workbook by clicking <a href="https://public.tableau.com/app/profile/ganesh.bharathwaj.v.g3124/viz/BARCELONASTATS/1VS1"><em>here</em></a></p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=600e20a21de2" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[SUPER RESOLUTION GAN]]></title>
            <link>https://msdganesh.medium.com/super-resolution-gan-adce6940b608?source=rss-2aee1599f169------2</link>
            <guid isPermaLink="false">https://medium.com/p/adce6940b608</guid>
            <category><![CDATA[artificial-intelligence]]></category>
            <category><![CDATA[machine-learning]]></category>
            <category><![CDATA[neural-networks]]></category>
            <dc:creator><![CDATA[Ganesh Bharathwaj]]></dc:creator>
            <pubDate>Fri, 27 Nov 2020 11:45:40 GMT</pubDate>
            <atom:updated>2020-11-27T12:08:06.725Z</atom:updated>
            <content:encoded><![CDATA[<p>We all love to see high resolution images but in reality we don’t see high resolution images all the time. Imagine you have designed a webpage like olx where your users will upload images of their products, we can’t guarantee that all images uploaded will be in good quality, sometimes we have to zoom our images to get more details from it but if we zoom, our image clarity gets distorted. This type disadvantages can be overcomed by Super resolution GAN(SRGAN). Before we get into SRGAN first we will understand what GAN is.</p><h3><strong>GAN</strong></h3><p>GAN stands for gnerative adversarial network which was developed by ian j Goodfellow in 2014. It consist of two neural networks which competes with each other and improves their output in each epoches.Those two neural networks are called Generator and Discriminator,Generator use to generate fake data like audio, video,etc to cheat the Discriminator and Discriminator distinguish between real and fake data, in this way both neural networks compete with each other and gets better and better in each epchoes.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/638/1*0Ukm_cwBltCALzmwgH6hCA.jpeg" /><figcaption>GAN architecture</figcaption></figure><h3>TYPES OF GAN</h3><p>1 Vanilla GAN</p><p>2 Conditional GAN</p><p>3 Deep Convolutional GAN</p><p>4 Laplacian Pyramid GAN</p><p>5 Super Resolution GAN</p><h3>SRGAN</h3><p>coming back to SRGAN, SRGAN uses GAN system to generate fake images with the help of generator and tries to fool Discriminator and Discriminator tries to differentiate between real data and fake data(generated by Genetrator). Each time when Generator gets caught by Discriminator, it improves its output and make sure that it generates an image which is very close to real data, on the other hand, discriminator improves its performance when it fails to identify the real data. A distorted image is sent as input to Generator and it generates a fake high resolution image which is very close to the real one so that it can fool Discriminator.By this way high resolution image is obtained from low resolution image. After many epoches genrator learns to create a perfect fake HR image from LR image.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*1SGIcLwdwuHvjaF1fjhnDQ.png" /><figcaption>SRGAN</figcaption></figure><p><strong>TYPES OF LOSS FUNCTION USED IN SRGAN</strong></p><p>1 Content loss</p><p>2 Adversarial Loss</p><p><strong>REFERENCES</strong></p><ul><li><a href="https://www.geeksforgeeks.org/super-resolution-gan-srgan/?ref=lbp">Super Resolution GAN (SRGAN) - GeeksforGeeks</a></li><li><a href="https://www.geeksforgeeks.org/generative-adversarial-network-gan/?ref=lbp">Generative Adversarial Network (GAN) - GeeksforGeeks</a></li></ul><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=adce6940b608" width="1" height="1" alt="">]]></content:encoded>
        </item>
        <item>
            <title><![CDATA[The necessity of weights and Bias in Neural Network]]></title>
            <link>https://msdganesh.medium.com/the-necessity-of-weights-and-bias-in-neural-network-dc520b67552a?source=rss-2aee1599f169------2</link>
            <guid isPermaLink="false">https://medium.com/p/dc520b67552a</guid>
            <dc:creator><![CDATA[Ganesh Bharathwaj]]></dc:creator>
            <pubDate>Thu, 13 Aug 2020 19:40:03 GMT</pubDate>
            <atom:updated>2020-08-13T19:40:03.989Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*7d10zUK4874FayT_AGCc2w.jpeg" /></figure><p>Hi, Usage of neural networks(NN) has been increased since the rapid increase of computational power and availability of enormous data on the internet. Weights and Bias plays a major role in NN but many of us don’t know the importance of these two parameters. In this blog let us see the importance of weights and basis.</p><p>We can understand something/someone importance only in their absence, so let us imagine that there is no such thing as weight or bias, and let&#39;s see what are all the consequences we would face if there is no such thing.</p><p>Weights help us to tell the orientation of the graph, It tells how steep the graph is. let us understand with an equation and graphs.</p><p>we know that<strong> Y=MX+C </strong>here M represents slope and C is Y-intersection on the graph. hypothesis function will look like <strong>h(x)=wx+b </strong>(similar to the previous equation)<strong> </strong>here Wis weight and B is bias. Let us imagine W is constant for different conditions.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*awPQJfFC3FJtOVO58bx24w.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*94mTQnAmtwEflDdaeRiyLg.png" /><figcaption><strong>GRAPH WITH SAME WEIGHT AND DIFFERENT BIAS</strong></figcaption></figure><p>From the above graph, we can infer that keeping weight constant does not change the steepness of the line, by just changing the bias we can only change the position of the line, not the orientation. what if we need a graph which looks like this</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/600/1*XF4vLJuLeWFjJ6Tzq3XtIA.png" /><figcaption><strong>Graph with different W</strong></figcaption></figure><p>can we achieve this graph with constant W? answer will be NO. so weight is a very important factor in neural networks.</p><h3>The role of weight in NN</h3><ol><li>weight determines the orientation of the graph</li><li>weight affects the output. For example, I want to purchase a motorcycle imagine I just considered only two features, Price and popularity. now my neural network should predict whether I buy this motorcycle or not(0-I won’t buy, 1- will buy). If I use the equation <strong>h(x)=w1x1+w2x2+b </strong>(x1-price, x2-popularity) and g(h(x)) as my activation function, to solve the problem with these equation I have certain drawbacks. If the price (x1) increases then my value of h(x) will increase eventually my g(h(x)) will also increase and I will get my output as 1 which means I will buy the motorcycle, in reality, people avoid motorcycle with a huge amount so to compensate this, weight comes into play. weight will have the highest negative number for a greater price. this makes the prediction right with highest -ve number h(x) will become -ve and finally, we get 0 as output.</li></ol><p>Now let&#39;s move to bias. similarly, we can keep bias constant and we change weight lets see what happens.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*-tM7JZw1ah59n6GF9K1dmA.png" /></figure><figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*ctzH8MRng5NEWMXG-jPm0Q.png" /><figcaption><strong>GRAPH WITH SAME BIAS</strong></figcaption></figure><p>we can clearly understand that changing W can only change the orientation of the graph, not the position. what if we need a graph which looks like this</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/800/1*414YjFg4CoF5pUQbx9IDsg.png" /><figcaption><strong>GRAPH WITH DIFFERENT BIAS</strong></figcaption></figure><p>Is that graph achievable with constant b? NO.</p><h3>Conclusion</h3><p>Bias and Weight are important they have a vital role in the Neural network. Hope this blog is helpful for you.</p><figure><img alt="" src="https://cdn-images-1.medium.com/max/1001/1*aswgXk3VHfYgZKlOHZ5XKQ.png" /></figure><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=dc520b67552a" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>