Impact of static and adaptive mutation techniques on the performance of Genetic Algorithm

Abstract

Genetic Algorithm GA is one of the most popular heuristic search algorithms inspired by nature's evolutionary behavior. Among the various genetic operators, mutation is one important operator that helps to accelerate the searching ability of GA. As GA finds numerous applications, it undergoes various enhancements and modifications, especially with respect to mutation operator. Numerous mutation techniques have been reported in the literature that can be broadly categorized into static and adaptive mutation techniques. This work selectively analyzes six mutation techniques in a common bench of experiments. Among the six mutation techniques, two are the popular variants of static mutation techniques called as Uniform mutation and Gaussian Mutation. The remaining four were recently introduced: two individual adaptive mutation techniques, a self adaptive mutation technique and a deterministic mutation technique. Totally, 28 benchmark functions, which fall under the benchmark categories of unimodal, multimodal, extended multimodal, diagonal and quadratic functions, are used in the work. The analysis mainly intends to determine a best mutation technique for every benchmark problem and to understand the dependency behavior of mutation techniques with other GA parameters such as crossover probabilities, population sizes and number of generations. It leads to interesting findings which would help to improve the GA performance on other practical and benchmark problems.