<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Knowledge Distillation on Nalar</title>
    <link>https://nalar.dev/tags/knowledge-distillation/</link>
    <description>Recent content in Knowledge Distillation on Nalar</description>
    <generator>Hugo</generator>
    <language>en-us</language>
    <lastBuildDate>Mon, 07 Sep 2026 00:00:00 +0700</lastBuildDate>
    <atom:link href="https://nalar.dev/tags/knowledge-distillation/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Compress Classifiers with Knowledge Distillation</title>
      <link>https://nalar.dev/compress-classifiers-with-knowledge-distillation/</link>
      <pubDate>Mon, 07 Sep 2026 00:00:00 +0700</pubDate>
      <guid>https://nalar.dev/compress-classifiers-with-knowledge-distillation/</guid>
      <description>&lt;p&gt;A model can be accurate enough for a product and still be too expensive to deploy. A large classifier may exceed a mobile memory budget, miss a latency target, or cost too much when every request requires substantial compute. Replacing it with a smaller model reduces those costs, but training the smaller model only from ground-truth labels can leave useful information behind.&lt;/p&gt;&#xA;&lt;p&gt;&lt;strong&gt;Knowledge distillation&lt;/strong&gt; addresses this problem by training a smaller &lt;strong&gt;student&lt;/strong&gt; model to learn from a stronger &lt;strong&gt;teacher&lt;/strong&gt; model. Instead of seeing only the correct class, the student can also learn how the teacher distributes its confidence across the alternatives.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
