Stochastic Local Search Using the Search Space Smoothing Meta-Heuristic: A Case Study

In this paper, two smoothing effects are firstly pointed out by analysis and by experiment on Traveling Salesman Problem(TSP) instances. We design a novel algorithm which runs stochastic local search under the SSS framework. The function determining the accepting probability of uphill moves is designed so that the algorithm can take advantage of the local smoothing effect ignored in original SSS. Experimental results on TSPLIB instances demonstrated that the performance of the new algorithm is much superior to traditional SSS approach.