混合性能指标优化问题的分阶段进化优化  

Phased Evolutionary Optimization for Problems with Hybrid Indices

作　　者：巩敦卫[1] 唐雷[2] 

机构地区：[1]中国矿业大学信息与电气工程学院,江苏徐州221008 [2]中国科学技术大学自动化系,安徽合肥230027

出　　处：《系统仿真学报》2007年第21期4936-4939,共4页Journal of System Simulation

基　　金：国家自然科学基金(60575046)

摘　　要：高维混合性能指标优化问题是普遍存在的,目前尚无有效的解决方法。提出一种解决该问题的分阶段进化优化方法。该方法根据性能指标的重要程度不同,将高维性能指标优化分为多个阶段,每一阶段只优化同层的性能指标;若性能指标间不是矛盾的,采用含权单目标进化优化方法求解;若性能指标间是矛盾的,采用Pareto进化优化方法求解。该方法可有效减轻用户疲劳,提高优化解的质量。在服装进化设计系统中的应用结果验证了方法的可行性和有效性。Optimization problems with multi-dimensional hybrid indices exist widely, and there is no effective method to solve them up to now. A phased evolutionary optimization algorithm for the problems above was proposed. Optimization with multi-dimensional hybrid indices was divided into several phases according to different important degrees of index. Only indices in the same layer were optimized in each phase. If there was no conflict among different indices, method for weighted single-objective evolutionary optimization was adopted; Otherwise, Pareto evolutionary optimization was adopted. The algorithm proposed could alleviate user fatigue and improve the quality of optimal solutions effectively. The results of application in fashion design validate its feasibility and efficiency.

关 键 词：进化优化 混合性能指标 高维 阶段 

分 类 号：TP301[自动化与计算机技术—计算机系统结构]