New Components with On-Line Control to Improve Harmony Search

Harmony Search is a new metaheuristic technique that uses fixed improvisation based on music to solve hard problems. We propose in this paper a new method to dynamically control and adapt the improvisation for Harmony Search. The key idea is to adapt the improvisation according to the search evaluation. Our solution provides best results at an affordable computational cost. We illustrate our proposal using the well-known Multidimentional Knapsack Problem, and compare our results with existing Harmony Search algorithms.