Files
myWiki/concepts/stein-lemma.md

29 lines
884 B
Markdown
Raw Blame History

This file contains ambiguous Unicode characters

This file contains Unicode characters that might be confused with other characters. If you think that this is intentional, you can safely ignore this warning. Use the Escape button to reveal them.

---
title: "Stein 引理"
created: 2026-06-22
updated: 2026-06-22
type: concept
tags: [mathematics, probability, statistics]
sources: [nano-filter]
---
# Stein 引理
Stein's lemma 是概率论中一个优雅的结果:对于联合高斯分布的随机变量,协方差可以通过期望的导数来计算。
## 基本形式
若 $(x, y)$ 服从联合高斯分布,且 $f$ 是几乎处处可微的函数,则:
$$
\text{Cov}(x, f(y)) = \text{Cov}(x, y) \cdot E[f'(y)]
$$
## 在 NANO 中的应用
[[nano-filter|NANO filter]] 利用 Stein 引理将 Bayesian 滤波中的预测步和更新步的最优性条件转化为解析形式。具体地Stein 引理允许将对数似然期望的梯度转化为可计算的矩表达式,从而将两个变分问题约化为有限维优化问题。
## 参考
- [[gaussian-filtering|Gaussian Filtering]]
- [[nano-filter|NANO Filter]]