A Double-Sampling Based Evolutionary Algorithm for Large-Scale Multi-Objective Optimization

Abstract

To speed up the convergence ability of the evolutionary algorithm, some direction-guided sampling techniques have been proposed for improving the search efficiency in large-scale multi-objective optimization. However, the approximate directions may not interact with the true Pareto set at all and lead to the ineffective sampling along the directions. In this paper, the double-sampling method in which the fuzzy Gaussian sampling is introduced to complement the direction-guided sampling is proposed to correct the search direction and improve the search efficiency. Moreover, a convergence-and-diversity-based mating selection is introduced to improve the distribution of the generated solutions. The experimental results on a variety of large-scale multi-objective problems with up to 5000 decision variables show the competitivity of the proposed algorithm compared with five state-of-the-art algorithms.

Keywords: Evolutionary multi-objective optimization, large-scale multi-objective problems, direction-guided sampling, reproduction