Performance Study of Double-Niched Evolutionary Algorithm on Multi-objective Knapsack Problems
Ryoya Osawa, Shinya Watanabe, Tomoyuki Hiroyasu, Satoru Hiwa · 2019
Multimodality is often observed in practical optimization problems. Therefore, multi-modal multi-objective evolutionary algorithms (MMEA) have been developed to tackle the multimodality of these problems. However, most of the existing studies focused on population diversity in either an objective or a decision space. A double-niched evolutionary algorithm (DNEA) is a state-of-the-art MMEA that employs a niche-sharing method to improve the population in both the objective and decision spaces. However, its performance has been evaluated solely for real-coded problems and not for binary-coded ones. In this study, the performance of DNEA is evaluated on a multi-objective 0/1 knapsack problem, and the population diversity in both the objective and decision spaces is evaluated using a pure diversity measure. The experimental results suggest that DNEA is effective for multi-objective 0/1 knapsack problems to improve the decision space diversity; further, its performance is significantly affected by its control parameter, niche radius.