Fine-grained tournament selection operator in genetic algorithms

Abstract

Tournament selection is one of the most popular selection operators in Genetic Algorithms. Recently, its popularity is increasing because this operator is well suited for Parallel Genetic Algorithms applications. In this paper, new selection operator is proposed. The new operator, which should be an improvement of the tournament selection, is named "Fine-grained Tournament Selection" (FGTS). It is shown that classical tournament selection is a special case of the FGTS and that new operator preserves its good features. Furthermore, theoretical estimations for the FGTS are made. Estimations for the FGTS are similar to those for the classical tournament selection. Finally, classical tournament selection, rank-based selection and FGTS are experimentally compared on a real world NP-hard problem and the obtained results are discussed.