<?xml version="1.0" encoding="utf-8" standalone="yes"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Optimization on Nalar</title>
    <link>https://nalar.dev/tags/optimization/</link>
    <description>Recent content in Optimization on Nalar</description>
    <generator>Hugo</generator>
    <language>en-us</language>
    <lastBuildDate>Fri, 04 Sep 2026 00:00:00 +0700</lastBuildDate>
    <atom:link href="https://nalar.dev/tags/optimization/index.xml" rel="self" type="application/rss+xml" />
    <item>
      <title>Understand Gradient Noise in Mini-Batch Training</title>
      <link>https://nalar.dev/understand-gradient-noise-in-mini-batch-training/</link>
      <pubDate>Fri, 04 Sep 2026 00:00:00 +0700</pubDate>
      <guid>https://nalar.dev/understand-gradient-noise-in-mini-batch-training/</guid>
      <description>&lt;p&gt;Neural network training usually updates model parameters from a small batch of examples rather than computing a gradient over the entire training set. That makes each update cheaper, but it also means the update direction depends on which examples happened to enter the batch.&lt;/p&gt;&#xA;&lt;p&gt;This variation is often called &lt;strong&gt;gradient noise&lt;/strong&gt;. It is not necessarily a bug. It is a consequence of estimating a dataset-wide gradient from a sample, and it creates an important trade-off between computation per update, update variability, and training throughput.&lt;/p&gt;</description>
    </item>
  </channel>
</rss>
