Mutual Information Invariance to Reparameterization

For two random variables X and Y, it is known that I(\phi(X);\phi(Y))=I(X,Y) which is also known as the reparameterization invariance property. To my known, the proof was given in Kraskov 2004 as follows

Let X'=\phi(X) and Y'=\phi(Y) denote the two mappings, respectively. The following equations hold

P(X',Y')=P(X,Y)|\frac{\partial X}{\partial X'}||\frac{\partial Y}{\partial Y'}|

and therefore 

I(X';Y')\\ =\int P(X',Y')\log \frac{P(X',Y')}{P(X')P(Y')}dX'dY' \\ =\int P(X,Y)\log \frac{P(X,Y)|\frac{\partial X}{\partial X'}||\frac{\partial Y}{\partial Y'}|}{P(X)|\frac{\partial X}{\partial X'}|P(Y)|\frac{\partial Y}{\partial Y'}|}dXdY \\ =\int P(X,Y)\log \frac{P(X,Y)}{P(X)P(Y)}dXdY\\ =I(X;Y)

  • 1
    点赞
  • 0
    收藏
    觉得还不错? 一键收藏
  • 0
    评论
评论
添加红包

请填写红包祝福语或标题

红包个数最小为10个

红包金额最低5元

当前余额3.43前往充值 >
需支付:10.00
成就一亿技术人!
领取后你会自动成为博主和红包主的粉丝 规则
hope_wisdom
发出的红包
实付
使用余额支付
点击重新获取
扫码支付
钱包余额 0

抵扣说明:

1.余额是钱包充值的虚拟货币,按照1:1的比例进行支付金额的抵扣。
2.余额无法直接购买下载,可以购买VIP、付费专栏及课程。

余额充值