504 Design of an Adaptive Search Space Smoothing Method and Experimental Analysis of its Search Characteristics
Kotaro Hiramatsu, Manabu Hasegawa · The Proceedings of Conference of Kanto Branch · 2012
An adaptive schedule for the smoothing parameter is introduced for the previously developed search space smoothing method in combination with the Metropolis algorithm (MAS3), and its performance is analyzed by a systematic numerical experiment on a random Euclidean traveling salesman problem. The results show that there exists an effective intermediate value of the smoothing parameter and that the performance is directly influenced by the total number of search steps and by the final value of the smoothing parameter: from the functionality point of view, the situation is quite the similar to the case of simulated annealing.