摘要： 为提高多目标进化算法的分布性和收敛性，提出一种基于海明距离差异的多目标进化算法。在非支配前沿的基础上定义海明等级，依据海明距离的大小对个体进行选择操作。同时结合海明差异和Pareto评价方法，对外部存储器中最优解进行更新和维护，通过结构相似度构建小生境空间，并引导算法趋向Pareto最优前沿面。对6个典型函数的测试结果表明，较其他对比算法，该算法在具备收敛性的同时能够保持较好的均匀性分布。

关键词: 多目标优化, 海明距离, 个体密度, 种群维护, 个体评价, Pareto最优

Abstract: In order to improve the distribution and convergence of multi-objective evolutionary algorithm, a kind of Hamming distance- based differences multi-objective evolutionary algorithm is proposed. Hamming grades are defined on the basis of the non-dominated frontier, choosing Hamming distance to operate the individuals. Hamming difference and Pareto evaluation methods are combined to update and maintain the optimal solution of the external memory, using structural similarity to build niche space and guide the algorithm towards Pareto optimal frontier. The test of 6 typical functions shows that, the proposed algorithm has better convergence while maintaining a good uniform distribution than other compared algorithms.

Key words: multi-objective optimization, Hamming distance, individual density, population maintenance, individual evaluation, Pareto optimization

中图分类号: 

• TP301.6