Fuzzy logic applied to tunning mutation size in evolutionary algorithms.

Tuning of parameters is a very important but complex issue in the Evolutionary Algorithms' design. The paper discusses the new, based on the Fuzzy Logic concept of tuning mutation size in these algorithms. Data on evolution collected in prior generations are used to tune the size of mutations. A Fuzzy Logic Part uses this historical data to improve the algorithm's convergence to a global optimum. The Fuzzy Logic Part keeps a desirable relation of exploration and exploitation, so the algorithm's resistance to getting stuck in a local optimum is improved too. Several tests on Function Optimization Problems were performed to prove the suitability of the proposed method. A set of data and functions with different difficulties, recommended in the commonly used benchmarks are used for experiments. The results of these experiments suggest that the proposed method is efficient and could be used for a wide range of similar problems of optimization.