混合性能指标优化问题的大种群规模进化算法

Evolutionary algorithm with large population size for problems with hybrid indices

中文关键词  进化优化  混合性能指标  遗传算法  交互

英文关键词  evolutionary optimization  hybrid indices  genetic algorithms  interaction

中文摘要

      混合性能指标优化问题可结合传统遗传算法和交互式遗传算法求解, 而种群规模和人机评价任务分配是影响算法性能的关键. 针对该问题, 本文提出一种新的进化优化算法. 首先, 采用大规模种群, 扩大搜索范围, 以增强算法的探索能力; 然后, 根据计算机和用户完成任务耗时的比值, 确定每代用户评价的个体数, 以提高计算机的使用效率; 接着, 采用K–均值聚类方法和基于相似度的估计策略, 以减轻用户疲劳; 最后, 采用Pareto占优比较不同个体的优劣, 使得最优解有较好的显式性能指标值和隐式性能指标值. 将本文算法应用于室内布局这一混合性能指标优化问题, 结果验证了所提算法的有效性.

英文摘要

      Problems with hybrid indices can be solved by combining traditional genetic algorithms and interactive genetic algorithms. The major causes affecting the performance of these algorithms are the population size and the strategy of task allocation between the user and the computer. A novel evolutionary optimization algorithm to solve the above problems is proposed. A large population is adopted to expand the searching space and enhance the performance of the algorithm in exploration. The number of individuals evaluated by the user in each generation is determined according to the ratio of the time consumptions of the computer and the user to finish their tasks, in order to improve the efficiency of the computer in usage. The K-mean clustering method and a similarity-based estimation strategy are adopted to alleviate the user’s fatigue. Finally, Pareto domination is employed to compare different individuals so that optimal solutions have good values in explicit and implicit indices. The proposed algorithm is applied to an interior layout problem with hybrid indices, and the results validate its efficiency.