Genetic algorithm for neural network architecture optimization

Abstract

The optimization of architecture of feed-forward neural networks is a complex task of high importance in supervised learning because it has a great impact on the convergence of learning methods. In this paper, we propose a multi-objective mathematical formulation in order to determine the optimal number of hidden layers, the number of neurons in each layer and good values of weights. We will solve our mathematical modeling using a hybridation of the famous genetic algorithm and the back-prop training algorithm. For evaluating our approach, we apply it to benchmark classification problems data-iris, seed, and wine. The obtained results demonstrated the effectiveness of the proposed approach.