New Genetic Algorithm for Neural Network Structure Optimization

Abstract

One of the very important tasks in modeling and control nonlinear processes is the determination of optimal artificial neural network (ANN) structure. Genetic algorithms (GA) are one of the many methods used for artificial neural networks structure optimization. Their disadvantage could be a long search time for finding of optimal solutions. This paper presents new algorithms based on genetic operations which enables to accelerate the optimization process using orthogonal activation function based neural networks (AOF NN).