在子集条件约束下的最大熵分布及其应用

Maximum Entropy Distribution Subject to Subset Constraints with Application

介绍了有关熵的概念及计算方法,并将其应用于构建一类新的分布估计算法(EDAs)。该类分布估计算法用基于最大熵估计种群中的模式概率分布和从最大熵分布中抽样取代遗传算法(GA)的交叉和变异,产生新的种群。在该类算法中,二阶连接模式算法由于只使用了连接模式,在解决变量之间相互作用趋向于发生在串中相互靠近的变量之间的一类问题时,比遗传算法更好。

After introducing some concepts and computations of entropy, a new type of estimation of distribution algorithms （EDAs） is developed by using principle of maximum entropy. This type of algorithms replaces the crossover and mutation operators used by genetic algorithm （GA） with the estimation of the maximum entropy distribution of schema in the population and sampling from maximum entropy distribution to generate new population. Among this type of algorithms, only contiguous schemata are used in order-2 contiguous schemata algorithm. Therefore, order-2 contiguous schemata algorithm may work better than GA when interactions between variables tend to be between variables that are located close to each other on the string.