Box-Muller变换：学习笔记

Haleine

已于 2022-07-20 08:24:39 修改

阅读量895

点赞数 3

分类专栏：笔记文章标签：学习算法

于 2022-07-18 21:22:53 首次发布

本文链接：https://blog.csdn.net/qq_30082785/article/details/125860498

版权

笔记专栏收录该内容

13 篇文章 5 订阅

订阅专栏

算法说明

Box-Muller变换是通过服从均匀分布的随机变量，来构建服从高斯分布的随机变量的一种方法。具体的描述为：选取两个服从 $[0, 1]$ 上均匀分布的随机变 $U_1$ 和 $U_2$ ，使得 $X$ ， $Y$ 满足高斯分布
$\sigma \cos (2\pi {U_1})\sqrt { - 2\ln {U_2}} + \mu., \quad Y = \sigma \sin (2\pi {U_1})\sqrt { - 2\ln {U_2}} + \mu.$
那么， $X$ ， $Y$ ~ $N(\mu , \sigma^2)$ 。

推导证明

$P roo f :$ Suppose $X$ and $Y$ satisfy $N(\mu, \sigma^2)$ , and they are $i . i . d .$ , the pdf is $p(X)=\frac{1}{\sqrt{2\pi} \sigma}e^{-\frac{(X-\mu)^2}{2\sigma^2}}, \quad p(Y)=\frac{1}{\sqrt{2\pi} \sigma}e^{-\frac{(Y-\mu)^2}{2\sigma^2}}.$
then we have
$P(Z)=P(XY)=P(X)P(Y)=\frac{1}{2\pi \sigma^2}e^{-\frac{(X-\mu)^2+(Y-\mu)^2}{2\sigma^2}}.$
Do coordinate transformation, then
$X=Rcos(\theta)+\mu, \quad Y=Rsin(\theta)+\mu.$
the sum of pdf of $Z$ can be written as
$1=\int\limits_{ - \infty }^\infty {\int\limits_{ - \infty }^\infty {\frac{1}{{2\pi {\sigma ^2}}}{e^{ - \frac{{{{(X - \mu )}^2} + {{(Y - \mu )}^2}}}{{2{\sigma ^2}}}}}dXdY} } =\int\limits_0^{2\pi } {\int\limits_0^\infty {\frac{1}{{2\pi {\sigma ^2}}}{e^{ - \frac{{{R^2}}}{{2{\sigma ^2}}}}}RdRd\theta } } .$
This leads to the distribution functions of $R$ and $θ$ are
$P(R\leq r)=\int\limits_0^{2\pi } d\theta {\int\limits_0^r {\frac{1}{{2\pi {\sigma ^2}}}{e^{ - \frac{{{R^2}}}{{2{\sigma ^2}}}}}RdR } } =1-e^{-\frac{r^2}{2\sigma^2}}, \quad P(\theta \leq \phi)==\int\limits_0^{\phi } d\theta{\int\limits_0^\infty {\frac{1}{{2\pi {\sigma ^2}}}{e^{ - \frac{{{R^2}}}{{2{\sigma ^2}}}}}RdR } }=\frac{\phi}{2\pi}.$
It is obvious that $\theta$ satisfies a uniform distribution on $2\pi]$ . Let
$F_R(r)=1-e^{-\frac{r^2}{2\sigma^2}},$
then its inverse function is
$R=F_R^{-1}(z)=\sigma\sqrt{-2ln(1-z)}.$
When $z$ satisfies the uniform distribution on $[0, 1]$ , the distribution function of R is $F_R(r)$ . So we can choose $U_1$ and $U_2$ which satisfy the uniform distribution, let
$\theta = 2\pi U_1,\quad R=\sigma\sqrt{-2lnU_2}.$
Bringing them to $X$ and $Y$ , then we can get that the initial $X$ or $Y$ satisfies the Gaussian distribution.