Files
myWiki/concepts/gaussian-manifold.md

30 lines
1.2 KiB
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: "高斯流形"
created: 2026-06-22
updated: 2026-06-22
type: concept
tags: [information-geometry, riemannian-geometry, gaussian-distribution]
sources: [nano-filter]
---
# 高斯流形
Gaussian manifold 是将高斯分布族视为 Riemannian 流形的几何框架。每个高斯分布 $N(x; \mu, \Sigma)$ 对应流形上的一点,参数 $v = (\mu, \text{vec}(\Sigma^{-1}))$ 作为坐标。
## Fisher 度量
在高斯流形上,自然度量是 Fisher 信息矩阵。对于参数化 $v$Fisher 矩阵的逆具有解析形式:
$$
F_v^{-1} = \begin{bmatrix} \Sigma & 0 \\ 0 & 2(\Sigma^{-1} \otimes \Sigma^{-1}) \end{bmatrix}
$$
## NANO 中的应用
[[nano-filter|NANO filter]] 的核心洞察Bayesian 滤波的更新步可以在高斯流形上视为一个优化问题——在流形上寻找一个高斯分布以最小化更新代价 $J(\hat{x}_t, P_t)$。由于高斯流形具有非平凡的曲率,标准梯度下降并非最陡下降方向;[[natural-gradient-descent|自然梯度下降]]用 Fisher 矩阵校正梯度,补偿流形的弯曲。
## 参考
- [[natural-gradient-descent|Natural Gradient Descent]]
- [[fisher-information-metric|Fisher Information Metric]]
- [[nano-filter|NANO Filter]]