Kullback Leibler Distance (or divergence)

 

K-L距离(相对熵)

K-L距离是两个完全确定的概率分布的非相似性度量。

1.定义:

p1(x)p2(x)是两个连续概率分布,根据定义,p1(x) p2(x)K-L距离D(p1,p2)为:

 

 

离散情况也类似。

如果用E表示p1分布的期望,上述表达式也可写作:

 

 

 

 

 

2.基本属性:

l         D(p1,p2) 的平均值,其中p1(x)为参考分布

l         K-L距离总是非负的。当两个分布想同时它的值是零。

l         K-L距离不是关于p1(x) p2(x)对称的,他也不是一个数学意义上的距离。

我们常常见到的是p1 p2K-L距离的的对称形式,即:

 

 

 

 

K-L距离通常很难精确地计算出来,但是两个分部都是正态分布是计算还是可以实现的。计算过程有点长但是相当有益的。

 

 

 

Kullback Leibler Distance (KL)

 
The Kullback Leibler distance (KL-distance) is a natural distance function from a "true" probability distribution, p, to a "target" probability distribution, q. It can be interpreted as the expected extra message-length per datum due to using a code based on the wrong (target) distribution compared to using a code based on the true distribution.
 
For discrete (not necessarily finite) probability distributions, p={p 1, ..., p n} and q={q 1, ..., q n}, the KL-distance is defined to be
 
KL(p, q) = Σi p i . log 2( pi / qi )
 
For continuous probability densities, the sum is replaced by an integral.
 
KL(p, p) = 0
KL(p, q) ≥ 0
 
Note that the KL-distance is not, in general, symmetric.

From:

http://www.aiaccess.net/English/Glossaries/GlosMod/e_gm_kullbak.htm#tut_1_basic

 

http://www.csse.monash.edu.au/~lloyd/tildeMML/KL/

 

 

你可能感兴趣的:(function,distance)