Combination of Guided Local Search and Estimation of Distribution Algorithm for Quadratic Assignment Problems

Guided Local Search (GLS) is a general meta-heuristic that sits on top of local search algorithms and help them escape from local minima. However, GLS is often trapped in a small local area. Estimation of Distribution Algorithm(EDA) is a new evolutionary algorithm in which statistical information are explicitly extracted from the current population and then used for generating new solutions. This paper combines EDA with GLS and proposes a hybrid algorithm, called EDA/GLS for Quadratic Assignment Problems (QAP). In EDA/GLS, a novel EDAlike operator, called guided mutation is used with GLS for exploitation. A restart strategy based on statistical information is employed for exploration. Experimental results show that EDA/GLS outperforms GLS.