科研之友
微信
新浪微博
Facebook
分享链接
摘要
Many multi-objective evolutionary algorithms (MOEAs) can converge to the Pareto optimal front and work well on two or three objectives, but they deteriorate when faced with many-objective problems. Indicator-based MOEAs, which adopt various indicators to evaluate the fitness values (instead of the Paretodominance relation to select candidate solutions), have been regarded as promising schemes that yield more satisfactory results than well-known algorithms, such as non-dominated sorting genetic algorithm (NSGA-II) and strength Pareto evolutionary algorithm (SPEA2). However, they can suffer from having a slow convergence speed. This paper proposes a new indicator-based multi-objective optimization algorithm, namely, the multiobjective shuffled frog leaping algorithm based on the epsilon indicator (epsilon-MOSFLA). This algorithm adopts a memetic meta-heuristic, namely, the SFLA, which is characterized by the powerful capability of global search and quick convergence as an evolutionary strategy and a simple and effective e-indicator as a fitness assignment scheme to conduct the search procedure. Experimental results, in comparison with other representative indicator-based MOEAs and traditional Pareto-based MOEAs on several standard test problems with up to 50 objectives, show that e-MOSFLA is the best algorithm for solving many-objective optimization problems in terms of the solution quality as well as the speed of convergence.
- 出版日期2014-8
- 单位深圳大学
