Dynamic regional harmony search with opposition and local learning

Harmony search (HS), mimicking the musician's improvisation behavior, has demonstrated strong efficacy in optimization. To deal with the deficiencies in the original HS, a dynamic regional harmony search (DRHS) algorithm with opposition and local learning is proposed. DRHS utilizes opposition-based initialization, and performs independent harmony searches with respect to multiple groups created by periodically regrouping the harmony memory. An opposition-based harmony creation scheme is used in DRHS to update each group memory. Any prematurely converged group is restarted with its size being doubled to enhance exploration. Local search is periodically applied to exploit promising regions around top-ranked candidate solutions. DRHS consistently outperforms HS on 12 numerical test problems from the CEC2005 benchmark at both 10D and 30D.