Investigation of Particle Multi-Swarm Optimization with Diversive Curiosity

This paper presents a new series of search methods of particle multi-swarm optimization (PMSO), which have intelligent judgment function in search process. The key idea, here, is first time systematically to create a psychological concept of diversive curiosity into the existing particle multi-swarm optimizers as an internal indicator. According to the idea, four search methods of PMSO with diversive curiosity, i.e. multiple particle swarm optimizers with information sharing and diversive curiosity (MPSOISDC), multiple particle swarm optimizers with inertia weight with information sharing and diversive curiosity (MPSOIWISDC), multiple canonical particle swarm optimizers with information sharing and diversive curiosity (MCPSOISDC), and hybrid particle swarm optimizers with information sharing and diversive curiosity (HPSOISDC) are proposed. This is a new technical expansion of PMSO in search framework for overcoming initial stagnation and avoiding boredom behavior to enhance search efficiency. In computer experiments, with adjusting the values of two parameters, i.e. duration of judgment and sensitivity, of the internal indicator, we inspect the performance index of the proposed methods by dealing with a suite of benchmark problems in search process. Based on detail analysis of the obtained experimental results, we reveal the outstanding search capabilities and characteristics of MPSOISDC, MPSOIWISDC, MCPSOISDC, and HPSOISDC, respectively.