Parameter Learning for Particle Swarm Optimization Using Deep Unfolding

Abstract

This paper proposes a tuning method of parameters included in particle swarm optimization (PSO) by utilizing deep unfolding. The proposed method provides the PSO algorithm with high convergence performance and accuracy by learning the parameters in advance. The proposed method is effective even when objective functions to be optimized change to a certain extent. Numerical experiments using several benchmark functions are conducted to demonstrate the effectiveness of the proposed method.