<?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>NADE on SummerFall's Blogs</title><link>https://summerfall1819.github.io/tags/nade/</link><description>Recent content in NADE on SummerFall's Blogs</description><generator>Hugo -- 0.145.0</generator><language>zh</language><copyright>Copyright © 2025 SummerFall</copyright><lastBuildDate>Thu, 09 Apr 2026 14:40:24 +0800</lastBuildDate><atom:link href="https://summerfall1819.github.io/tags/nade/index.xml" rel="self" type="application/rss+xml"/><item><title>自回归模型</title><link>https://summerfall1819.github.io/posts/generativemodel/chap2/</link><pubDate>Thu, 09 Apr 2026 14:40:24 +0800</pubDate><guid>https://summerfall1819.github.io/posts/generativemodel/chap2/</guid><description>&lt;p>自回归模型（autoregressive model, AR）是最直接的一类显式生成模型。它并不试图一次性写出整个高维联合分布，而是先给随机变量规定一个顺序，再把联合分布拆成一串条件分布。这样做的结果是：模型既可以精确计算样本概率，又可以按顺序采样生成新样本。因此，自回归模型长期是密度估计、语言模型与图像生成中的核心方法。&lt;/p></description></item></channel></rss>