基于最大熵的分布估计算法  

A Framework for Estimation of Distribution Algorithms Based on Maximum Entropy

作　　者：姜群[1] 王越[1] 

机构地区：[1]重庆工学院计算机学院,重庆400050

出　　处：《微电子学与计算机》2007年第11期73-76,共4页Microelectronics & Computer

基　　金：重庆市自然科学基金项目(CSTC2006BB2397)

摘　　要：分布估计算法是当前进化计算领域的一个新方向。文中提出一种新的基于最大熵的分布估计算法,主要用基于最大熵估计种群中的模式概率分布,取代贝叶斯网络分布估计算法中的贝叶斯概率图模型。该算法无需进行贝叶斯网络学习,大大减少了计算量,而且还能获取更准确的概率分布估计。实验结果表明,与贝叶斯优化算法相比,该算法具有更高的稳定性和更强的寻优能力。Estimation of Distribution Algorithms is a new area of evolutionary computation. In this paper, a framework for a new type of estimation of distribution algorithms（EDA） based on Maximum Entropy was developed. The new algorithms are similar to the Bayesian Optimization Algorithm except that they replace Bayesian network model with estimation of schema distribution based on maximum entropy. As structure learning of Bayesian network is not needed, they reduce the computational cost. Furthermore, the better estimation of the joint probability distribution can be obtained. The experimental results show that the new algorithms achieve more stable performance and stronger ability in searching the global optima.

关 键 词：熵 模式 概率分布 约束 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]