<?xml version="1.0" encoding="utf-8" standalone="yes"?><rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom" xmlns:content="http://purl.org/rss/1.0/modules/content/"><channel><title>Pytorch on Sanketh's Blog</title><link>https://sankethbk.github.io/blog/tags/pytorch/</link><description>Recent content in Pytorch on Sanketh's Blog</description><generator>Hugo -- 0.170.1</generator><language>en-us</language><lastBuildDate>Sat, 21 Nov 2026 00:00:00 +0530</lastBuildDate><atom:link href="https://sankethbk.github.io/blog/tags/pytorch/index.xml" rel="self" type="application/rss+xml"/><item><title>Lab 1.4: Weight Sharing — From Position-Anchored Parts to Convolutional Features</title><link>https://sankethbk.github.io/blog/posts/ml/labs/2026-11-21-lab-4-weight-sharing-and-representations/</link><pubDate>Sat, 21 Nov 2026 00:00:00 +0530</pubDate><guid>https://sankethbk.github.io/blog/posts/ml/labs/2026-11-21-lab-4-weight-sharing-and-representations/</guid><description>&lt;h1 id="lab-14-weight-sharing--from-position-anchored-parts-to-convolutional-features"&gt;Lab 1.4: Weight Sharing — From Position-Anchored Parts to Convolutional Features&lt;/h1&gt;
&lt;p&gt;&lt;strong&gt;MNIST internals series.&lt;/strong&gt; &lt;a href="https://sankethbk.github.io/blog/posts/ml/labs/2026-10-19-lab-1-mnist-from-pixels-to-neural-network/"&gt;Lab 1.1&lt;/a&gt; gave us class templates. &lt;a href="https://sankethbk.github.io/blog/posts/ml/labs/2026-10-19-lab-2-continual-learning-1-2-then-3/"&gt;Lab 1.2&lt;/a&gt; watched them reorganize. &lt;a href="https://sankethbk.github.io/blog/posts/ml/labs/2026-10-27-lab-3-from-templates-to-features/"&gt;Lab 1.3&lt;/a&gt; added a hidden layer and the templates dissolved into shared, position-anchored stroke fragments.&lt;/p&gt;
&lt;p&gt;That last phrase is the loose thread. A fully connected hidden unit has a &lt;strong&gt;different weight for every pixel&lt;/strong&gt;, so detecting the same short stroke at two locations costs two different sets of weights. Nothing in the architecture says that a useful pattern in the top-left corner should also be useful in the bottom-right.&lt;/p&gt;</description></item></channel></rss>