GENETIC ALGORITHM PARAMETER TUNING USING EXPLORATORY LANDSCAPE ANALYSIS AND MACHINE LEARNING

Abstract

The choice of parameter values in evolutionary algorithms greatly affects their performance. Many popular parameter tuning methods are constrained by the maximum number of fitness function evaluations to find a good set of parameter values. Recently, an approach to algorithm selection for optimization problems has been proposed, which uses the analysis of the fitness function landscape and machine learning to select the optimal algorithm based on the characteristics of its landscape. Such application of fitness landscape analysis motivates further research, particularly in the context of parameter tuning in evolutionary algorithms. The use of landscape features allows for the identification of similar problems and the use of parameter tuning data obtained from testing on benchmark problems, significantly reducing the number of required fitness function evaluations during tuning. This work considers an approach to automatic parameter selection using landscape analysis of the objective function and machine learning, using a genetic algorithm as an example. The proposed solution evaluates the characteristics of the landscape of the optimization problem's objective function and suggests optimal parameter values for the algorithm using a neural network. This network was trained on a dataset of landscape features expressed as numerical features and their corresponding optimal algorithm parameter sets. In contrast to approaches for automatic algorithm selection for a specific problem, this work addresses the problem of regressing algorithm parameters instead of classifying the most suitable algorithm from a given set. The results of experiments on different configurations of the W-model problem, as well as on the MAX-3SAT problem, show that the proposed approach to automatic parameter selection considering the landscape of the objective function can help determine appropriate values for the static parameters of the genetic algorithm. The algorithm with the proposed parameter values outperforms other considered options on average, requiring fewer evaluations of the objective function to find the optimum compared to the other algorithms considered.