A self-adaptive heterogeneous pso for real-parameter optimization

Abstract

Heterogeneous particle swarm optimizers (HPSO) allow particles to use different update equations, referred to as behaviors, within the swarm. Dynamic HPSOs allow the particles to change their behaviors during the search. These HPSOs alter the exploration/exploitation balance during the search which alters the search behavior of the swarm. This paper introduces a new self-adaptive HPSO and compares it with other HPSO algorithms on the CEC 2013 real-parameter optimization benchmark functions. The proposed algorithm keeps track of how successful each behavior has been over a number of iterations and uses that information to select the next behavior of a particle. The results show that the proposed algorithm outperforms existing HPSO algorithms on the benchmark functions.