The property of Gaussian

「已注销」

于 2018-08-09 08:58:29 发布

阅读量264

点赞数

分类专栏：随笔

本文链接：https://blog.csdn.net/u012284960/article/details/81530168

版权

随笔专栏收录该内容

10 篇文章 1 订阅

订阅专栏

Assume $\boldsymbol{x} \sim \mathcal{N}(\boldsymbol{a},\boldsymbol{A})$ , prove that

P x \sim N (H x, P A P T) (266)

$\begin{align} \boldsymbol{P}\boldsymbol{x}\sim \mathcal{N}(\boldsymbol{Hx}, \boldsymbol{P} \boldsymbol{A}\boldsymbol{P}^T) \end{align}$
where

x∈Rn x ∈ R n $\boldsymbol{x}\in \mathbb{R}^n$ ,

P∈Rm×n P ∈ R m × n $\boldsymbol{P}\in \mathbb{R}^{m\times n}$ . Notation

N(a,A) N ( a , A ) $\mathcal{N}(\boldsymbol{a},\boldsymbol{A})$ denotes a Gaussian distribution with

a a $\boldsymbol{a}$ mean and

A A $\boldsymbol{A}$ covariance martix. In addition,

(⋅)T ( ⋅ ) T $(\cdot)^T$ refers to transport operation such as

(P)T ( P ) T $(\boldsymbol{P})^T$ .

Proof : Let $\boldsymbol{y}=\boldsymbol{Px}$ , then its CHF ( characteristic function ) is

ϕ y (w) = E [exp (j w T y)] = E [exp (j (P T w) T x)] = exp (j (P T w) T a - 1 2 (P T w) T A P T w) = exp (j w T (P a) - 1 2 w T (P A P T) w) (267) (268) (269) (270)

$\begin{align} \phi_{\boldsymbol{y}}(\boldsymbol{w})&=\mathbb{E}\left[{\exp(j\boldsymbol{w}^T\boldsymbol{y})}\right]\\ &=\mathbb{E}\left[{\exp(j(\boldsymbol{P}^T\boldsymbol{w})^T\boldsymbol{x})}\right]\\ &=\exp\left({j(\boldsymbol{P}^T\boldsymbol{w})^T\boldsymbol{a}-\frac{1}{2}(\boldsymbol{P}^T\boldsymbol{w})^T\boldsymbol{A P}^T\boldsymbol{w}}\right)\\ &=\exp\left({j\boldsymbol{w}^T(\boldsymbol{P a})-\frac{1}{2}\boldsymbol{w}^T(\boldsymbol{P A P}^T)\boldsymbol{w}}\right) \end{align}$
Note that the CHF of

x x $\boldsymbol{x}$ is

exp(jwTa−12wTAw) exp ⁡ ( j w T a − 1 2 w T A w ) $\exp\left({j\boldsymbol{w}^T\boldsymbol{a}-\frac{1}{2}\boldsymbol{w}^T\boldsymbol{A}\boldsymbol{w}}\right)$ . Hence, the mean and variance of

y y $\boldsymbol{y}$ are

Pa P a $\boldsymbol{Pa}$ and

PAPT P A P T $\boldsymbol{PA P}^T$ , respectively.

「已注销」

关注

0
点赞
踩
0

收藏

觉得还不错? 一键收藏
0
评论
The property of Gaussian

Assume x∼N(a,A)x∼N(a,A)\boldsymbol{x} \sim \mathcal{N}(\boldsymbol{a},\boldsymbol{A}), prove that Px∼N(Hx,PAPT)(266)(266)Px∼N(Hx,PAPT)\begin{align}\boldsymbol{P}\boldsymbol{x}\sim \mathcal{N}(\bold...
复制链接

扫一扫