A hybrid harmony search algorithm combined with differential evolution for global optimization problems

Harmony search (HS) is a recently proposed meta-heuristic by imitating music improvisation process, which has drawn much attention in the past few years. However, researches have revealed that the performance and the convergence rate of the method are suffered when dealing with high-dimensional or/and multimodal problems. To get a better control between exploitation and exploration, a hybrid HS algorithm is proposed, which is characterized in two aspects. First, the memory consideration scheme is modified by introducing crossover and mutation operators, which is inspired by the differential evolution (DE) algorithm. Second, two control parameters, namely PAR and bw, are either dynamically adjusted or self-learning along with the evolution process to fine-tune the solutions. Numerical results based on a test suite of well-known benchmark functions show that the proposed algorithm is more effective or at least competitive in finding near-optimal solutions compared with three HS variants and the DE/rand/1/bin algorithm.