An adaptive human learning optimization with enhanced exploration–exploitation balance

Abstract

Human Learning Optimization (HLO) is a simple yet efficient binary meta-heuristic, in which three learning operators, i.e. the random learning operator (RLO), individual learning operator (ILO) and social learning operator (SLO), are developed to mimic human learning mechanisms to solve optimization problems. Among these three operators, RLO directly influences the exploration and exploitation abilities of HLO, and therefore its control parameter pr is of great importance since it controls the balance between exploration and exploitation. In this paper, an adaptive human learning optimization with enhanced exploration-exploitation balance (AHLOee) is proposed to improve the performance of HLO, in which a new adaptive pr strategy is carefully designed to meet the different requirements of HLO at different stages of iterations. A comprehensive parameter study is performed to evaluate the influences of the proposed adaptive strategy on exploration and exploitation, and then the deep insights on the role of RLO and the reason why the proposed adaptive strategy can achieve a practically ideal trade-off between exploration and exploitation are provided. The experimental results on the CEC05 and CEC15 benchmarks demonstrate that the proposed AHLOee has advantages over previous HLO variants and outperforms recent state-of-art binary meta-heuristics.