紀要論文

Adaptive Update of Reference Point Value for Parallel and Distributed MOEA/D

eng

24321192

https://doi.org/10.15002/00026286

法政大学大学院情報科学研究科

法政大学大学院紀要. 情報科学研究科編

18

1

6

2023-03-24

Version of Record

This paper proposes an updated method to determine the reference point for acceleration based on Multi-Objective Evolutionary Optimization Algorithm Parallel and Distributed Decomposition (MOEA/D) to process a multi-core environment. A method had been already proposed in switching update intervals which is not only effective for decreasing sparse regions in the initial stage, and improves multiformity and increases comparison of HV value and virtual overlap region method. However, this proposed method has limitations because the switching of the update interval needs to be manually set to find an appropriate switching value. In this paper, a method for automatically switching the update interval is proposed, which can be applied to various problems. Evaluating a newly proposed method, the convergence and diversity were compared by using a single-core MOEA/D and a parallel MOEA/D constrained knapsack matter, and prove that with a small number of generations it is not only effective for reducing the sparse regions up and also can improve diversity and increase HV value.

Article

AA12746425