Adaptive Tabu Tenure Computation in Local Search

Optimization methods based on complete neighborhood exploration such as Tabu Search are impractical against large neighborhood problems. Strategies of candidate list propose a solution to reduce the neighborhood exploration complexity. We propose in this paper a generic Tabu Search algorithm using adaptive candidate list strategy based on two alternate candidate lists. Each candidate list strategy corresponds to a given search phase: intensification or diversification. The optimization algorithm uses a Tabu list containing the variables causing loops. The paper proposes a classification of Tabu tenure managing in the literature and presents a new and original Tabu tenure adaptation mechanism. The generic method is tested on the k-coloring problem and compared with some best methods published in the literature. Obtained results show the competitiveness of the method.