A New Differential Mutation Based Adaptive Harmony Search Algorithm for Global Optimization

Abstract

The canonical harmony search (HS) algorithm generates a new solution by using random adjustment. However, the beneficial effects of harmony memory are not well considered. In order to make full use of harmony memory to generate new solutions, this paper proposes a new adaptive harmony search algorithm (aHSDE) with a differential mutation, periodic learning and linear population size reduction strategy for global optimization. Differential mutation is used for pitch adjustment, which provides a promising direction guidance to adjust the bandwidth. To balance the diversity and convergence of harmony memory, a linear reducing strategy of harmony memory is proposed with iterations. Meanwhile, periodic learning is used to adaptively modify the pitch adjusting rate and the scaling factor to improve the adaptability of the algorithm. The effects and the cooperation of the proposed strategies and the key parameters are analyzed in detail. Experimental comparison among well-known HS variants and several state-of-the-art evolutionary algorithms on CEC 2014 benchmark indicates that the aHSDE has a very competitive performance.