zoukankan      html  css  js  c++  java
  • MATLAB小函数:计算KL散度与JS散度

    MATLAB小函数:计算KL散度与JS散度

    作者:凯鲁嘎吉 - 博客园 http://www.cnblogs.com/kailugaji/

    问题:给定两个向量,计算这两个向量之间的Kullback-Leibler Divergence与Jensen-Shannon Divergence。KL散度与JS散度的计算公式参考:相似性度量 - 凯鲁嘎吉 - 开发者的网上家园 

    1. MATLAB程序

    function [score_KL, score_JS] = KL_JS_div(vec1, vec2)
    % Input: vec1: vector 1, vec2: vector 2
    % Output: score_KL: KL divergence, source_JS: JS divergence
    % Author: kailugaji 
    % https://www.cnblogs.com/kailugaji/
    
    % Make sure vec1 and vec2 sum to 1
    if any(vec1(:))
        vec1 = vec1/sum(vec1(:));
    end
     
    if any(vec2(:))
        vec2 = vec2/sum(vec2(:));
    end
    
    % Compute Kullback-Leibler Divergence
    score_KL = sum(sum(vec1.* log(eps + vec1./(vec2+eps))));
    
    % Compute Jensen-Shannon Divergence
    score_JS = (sum(sum(vec1.* log(eps + vec1./((vec1+vec2)./2+eps))))+sum(sum(vec2.* log(eps + vec2./((vec1+vec2)./2+eps)))))./2;
    
    if vec1==vec2
        score_KL=0;
        score_JS=0;
    end

    2. 结果

    >> vec1=[0.2 0.4 0.4];
    >> vec2=[0.3 0.2 0.5];
    >> [score_KL, score_JS] = KL_JS_div(vec1, vec2)
    
    score_KL =
    
       0.106908430076661
    
    
    score_JS =
    
       0.024807303850391
    

    2020-09-30

  • 相关阅读:
    ISBN号码
    计数问题
    小玉在游泳
    数字反转
    单调队列(学习笔记)
    LCA(学习笔记)
    emacs配置文件
    线段树(学习笔记)
    RMQ问题 ST算法(学习笔记)
    Lucas卢卡斯定理(学习笔记)
  • 原文地址:https://www.cnblogs.com/kailugaji/p/13755159.html
Copyright © 2011-2022 走看看