A niching framework based on fitness proportionate sharing for multi-objective genetic algorithm (MOGA-FPS)

Authors:

Abstract
In this paper, we propose a new niching framework based on Fitness Proportionate Sharing (FPS) to improve the diversity preservation in multi-objective optimization. The traditional sharing approach in standard Multi-Objective Genetic Algorithm (MOGA) is replaced with the proposed niching framework and the adapted MOGA is named MOGA-FPS. We also propose an algorithm which dynamically finds the most suitable niche radius. Experimental results show that MOGA-FPS significantly improves MOGA performance and maintains a well spread distribution of optimal solution set for bi-objective test functions compared with NSGA-II, MOGAS (Multi-Objective Genetic Algorithm using a new fitness sharing function).
Funding Information
  • Undersecretary of defense (FA8750-15-2-0116)