GOLS—Genetic orthogonal least squares algorithm for training RBF networks

Abstract

This work presents the genetic orthogonal least squares approach, a hybrid algorithm blending orthogonal least squares method with a genetic algorithm at the same hierarchical level. The resulting method assimilates the advantages of both original approaches, generating solutions substantially better than those produced by the orthogonal least squares algorithm, without incurring in the computational cost of the standard genetic algorithm. To support this statement, the new approach is submitted to several computational experiments with artificial and real-world data.