Fuzzy logic applied to mutation size in evolutionary strategies

Tuning of algorithm parameters is a complex but very important issue in the design of Evolutionary Algorithms. This paper discusses a new concept of mutation size tuning in Evolutionary Strategies. The proposed algorithm uses data on evolutionary history in earlier generations to tune the mutation size. A Fuzzy Logic Part examines this historical data and tunes the mutation size of individuals to improve the algorithm's convergence and its resistance to getting stuck in a local optimum. The Fuzzy Logic Part tunes the mutation size and keeps an appropriate relation of algorithm's exploration and exploitation. The proposed concept is discussed, and several tests on Function Optimization Problems are performed. In tests, we use a set of data and functions with different difficulties recommended in the commonly used benchmarks. The results of experiments suggest that the proposed method is more efficient and resistant to getting stuck in suboptimal solutions. The proposed algorithm has been used in recognizing the type of ultra-high energy cosmic ray particle that initiates the Extensive Air Showers when hit the Earth atmosphere. It could be used for a wide range of similar problems. It is possible that the proposed method could be adapted to other types of optimization methods, inspired by natural evolution, for example, Evolutionary Algorithms.