Multi-chaotic system induced success-history based adaptive differential evolution

Abstract

This research paper combines two soft computing fields – chaos theory and evolutionary computing. The proposed multi-chaotic system implements five different chaotic maps as a Pseudo-Random Number Generators (PRNGs) for parent selection process in Differential Evolution (DE) and Success-History based Adaptive Differential Evolution (SHADE) algorithms. The probabilities for selecting chaotic maps are adapted and the adaptation process is based on the previous successful solutions. Therefore, PRNG varies for different test functions. The performance of multi-chaotic system induced DE and SHADE is compared against their canonical versions on CEC2015 benchmark set. Acquired results show that replacing classic PRNG with multi-chaotic PRNG can lead sto improvement in terms of convergence speed and ability to reach the global optimum.