InTech-Recent_advances_in_harmony_search外文书.pdf

7 Recent Advances in Harmony Search Zong Woo Geem1, M. Fesanghary2, Jeong-Yoon Choi3, M. P. Saka4, Justin C. Williams1, M. Tamer Ayvaz5, Liang Li6, Sam Ryu7 and A. Vasebi8 1Johns Hopkins University, 2Amirkabir University of Technology, 3Montgomery College, 4Middle East Technical University, 5Pamukkale University, 6Da Lian University of Technology, 7SOFEC, 8K.N.Toosi University of Technology 1,3,7USA, 2,8Iran, 4,5Turkey, 6China 1. Introduction The harmony search (HS) is a music-inspired evolutionary algorithm, mimicking the improvisation process of music players (Geem et al., 2001). The HS is simple in concept, few in parameters, and easy in implementation, with theoretical background of stochastic derivative (Geem, 2007a). The algorithm was originally developed for discrete optimization and later expanded for continuous optimization (Lee Geem, 2005). The following pseudo code describes how the HS algorithm works: procedure HS // initialize initiate parameters initialize the harmony memory //main loop while (not_termination) for I = 1 to number of decision variables (N) do R1 = uniform random number between 0 and 1 if (R1 P HMCR ) (memory consideration) X[I] will be randomly chosen from harmony memory R2 = uniform random number if (R2 P PAR ) (pitch adjustment) X[I] = X[I] ± ? end if else (random selection) X[I] = X ∈ Φ (Φ = Value Set) end if end do // evaluate the fitness of each vector fitness_X = evaluate_fitness(X) // update harmony memory update_memory(X, fitness_X) % if applicable end while end procedure Advances in Evolutionary Algorithms 128 Ensemble harmony search (EHS) is another variant of the HS where ensemble consideration is added to the original algorithm structure (Geem, 2006a). The new operation considers the relationship among decision variables. The EHS could overcome the drawback