进化高维多目标优化算法研究综述

Survey on evolutionary many-objective optimization algorithms

首先针对常规多目标优化算法求解高维多目标优化时面临的选择压力衰减问题进行论述;然后针对该问题,按照选择机制的不同详细介绍基于Pareto支配、基于分解策略和基于性能评价指标的典型高维多目标优化算法,并分析各自的优缺点;接着立足于一种全新的性能评价指标—–R2指标,给出R2指标的具体定义,介绍基于R2指标的高维多目标优化算法,分析此类算法的本质,并按照R2指标的4个关键组成部分进行综述;最后,发掘其存在的潜在问题以及未来发展空间.

The paper first presents a comprehensive description of the selection pressure degradation problem when the traditional multi-objective evolutionary algorithms are adopted to solve many-objective optimization problems.According to the selection strategies, the Pareto-based, decomposition-based and indicator-based many-objective evolutionary algorithms（Ma OEAs） are proposed to address the selection pressure degradation problem. Then, the definition of R2 indicator proposed as a new performance indicator is provided, and a series of R2-Ma OEAs are infroduced. After analyzing the nature of R2-Ma OEAs, this paper gives a general review of these algorithms in consideration of their four main components. Finally, the inherent basis and the future research arevintroduced.