<?xml version="1.0" encoding="UTF-8"?>
<rss version="2.0" xmlns:atom="http://www.w3.org/2005/Atom">
  <channel>
    <title>Amir Ghazizadeh — Blog</title>
    <link>https://amir-ghz.github.io/blog/</link>
    <atom:link href="https://amir-ghz.github.io/feed.xml" rel="self" type="application/rss+xml"/>
    <description>Sparse acceleration, GPU kernels, and hardware-software co-design.</description>
    <language>en-us</language>
    <lastBuildDate>Sun, 31 Aug 2026 00:00:00 +0000</lastBuildDate>
    <item>
      <title>Two lotteries, and the space between them</title>
      <link>https://amir-ghz.github.io/blog/two-lotteries/</link>
      <guid isPermaLink="true">https://amir-ghz.github.io/blog/two-lotteries/</guid>
      <pubDate>Sun, 31 Aug 2026 00:00:00 +0000</pubDate>
      <author>amir.g@ucf.edu (Amir Ghazizadeh)</author>
      <category>Sparsity</category>
      <category>Hardware accelerators</category>
      <description>Machine learning made two lottery arguments seven years apart and never put them in the same room. One says most of a trained network is surplus. The other says the hardware of the day decides whether you can collect on it.</description>
    </item>
  </channel>
</rss>
