A comparative study of fuzzy parameter control in a general purpose local search metaheuristic
Jackson, Warren G. and Özcan, Ender and John, Robert I. (2016) A comparative study of fuzzy parameter control in a general purpose local search metaheuristic. In: 2016 IEEE Congress on Evolutionary Computation (CEC), 24-29 July, 2016, Vancouver, Canada. (In Press)
There is a growing number of studies on general purpose metaheuristics that are directly applicable to multiple domains. Parameter setting is a particular issue considering that many of such search methods come with a set of parameters to be configured. Fuzzy logic has been used extensively in control applications and is known for its ability to handle uncertainty. In this study, we investigate the potential of using fuzzy systems to control the parameter settings of a threshold accepting (TA) metaheuristic for improving the overall effectiveness of a cross-domain approach. We have evaluated the performance of various general purpose local search metaheuristics which mix multiple heuristics at random and apply the TA metaheuristic with fixed threshold, crisp (non-fuzzy) rule-based control of the threshold and various fuzzy systems controlling the threshold. The empirical results show that the approach using the TA with crisp rule-based control performs the best across six problem domains from a benchmark.
Actions (Archive Staff Only)