Opposition-based learning in the shuffled bidirectional differential evolution algorithm

Abstract

The opposition-based learning (OBL) strategy by comparing the fitness of an individual to its opposite and retaining the fitter one in the population accelerates search process. In this paper, the OBL is employed to speed up the shuffled bidirectional differential evolution (SBDE) algorithm. The SBDE by employing the partitioning, shuffling and bidirectional optimization concepts increases the number and diversity of search moves in respect to the original differential evolution (DE). So with incorporating the SBDE and OBL strategy, we can obtain the algorithms with an ability of better exploring the promising areas of search space without occurring stagnation or premature convergence. Experiments on 25 benchmark functions and non-parametric analysis of obtained results demonstrate a better performance of our proposed algorithms than original SBDE algorithm. Also an extensive performance comparison the proposed algorithms with some modern and state-of-the-art DE algorithms reported in the literature confirms a statistically significantly better performance of proposed algorithms in most cases. In a later part of the comparative experiments, firstly proposed algorithms are compared with other evolutionary algorithms (EAs) proposed for special session CEC2005. Then a comparison against a wide variety of recently proposed EAs is performed. The obtained results show that in most cases the proposed algorithms have a statistically significantly better performance in comparable to several existing EAs.