A New Algorithm to Design Multiple Hidden Layered Artificial Neural Networks

Abstract

This paper presents a new constructive algorithm known as multilayered constructive architecture (MCA) for designing and training multiple hidden layered artificial neural networks (ANNs). Unlike most previous constructive algorithms, MCA puts emphasis on both simplicity and generalization ability of designed ANNs. In order to maintain simplicity, both the number of hidden layers and nodes in a hidden layer are determined by a constructive approach. The use of modified single layer backpropagation algorithm for training output layer and hidden neurons increases the generalization ability of the ANN. MCA has been tested extensively on a number of benchmark problems in machine learning and neural networks, including Australian credit card assessment, breast cancer, diabetes, glass and heart disease. The experimental results show that MCA can produce compact ANNs with good generalization ability.