Department of Applied Mathematics & Physics, Kyoto University

Technical Report 2003-007 (June 21, 2003)

Tabu Search Directed by Direct Search Methods for Nonlinear Global Optimization
by Abdel-Rahman Hedar and Masao Fukushima

PostScript File

In recent years, there has been a great deal of interest in metaheuristics in the optimization community. Tabu Search (TS) represents a popular class of metaheuristics. However, compared with other metaheuristics like genetic algorithm and simulated annealing, contributions of TS that deal with continuous problems are still very limited. In this paper, we introduce a continuous TS called Directed Tabu Search (DTS) method. In the DTS method, direct-search-based strategies are used to direct a tabu search. These strategies are based on the well-known Nelder-Mead method and a new pattern search procedure called adaptive pattern search. Moreover, we introduce a new tabu list conception with anti-cycling rules called Tabu Regions and Semi-Tabu Regions. In addition, Diversification and Intensification search schemes are employed. Numerical results show that the proposed method is promising and produces high quality solutions.