构建在最大熵原理上的分布估计算法及其应用研究

Constructing Estimation of Distribution Algorithms Based on the Maximum Entropy

分布估计算法是进化计算领域的一个新方向.它主要用概率模型来建造进化计算中的遗传算法,它不再依赖于交叉与变异,而是估计较好个体的概率分布,用概率分布来引导对搜寻空间的探索.本文提出一类基于最大熵的分布估计算法.实验结果表明,在解决某些较复杂问题时,本文算法比遗传算法更具优势.

Estimation of Distribution Algorithm （EDA） is a new area of Evolutionary Computation. Unlike traditional Genetic Algorithms, EDAs replace the crossover and mutation operators by constructing a probabilistic model of promising individuals, and exploit search space guided by the probability distribution. This paper develops a framework for estimation of distribution algorithms based on maximum entropy. Empirical results show that an algorithm of this type gives better performance than genetic algorithm in some complex problems.