New Methods of Training Two-Layer Sigmoidal Neural Networks with Regularization

Abstract

We propose an algorithm for training two-layer sigmoidal artificial neural networks (ANN) in the presence of significant interference and low-informative variables. To obtain an efficient initial ANN parameters approximation, the algorithm applies a uniform distribution of the neuron work areas in the data domain, followed by training of fixed neurons. The proposed algorithm for ANN learning in combination with non-smooth regularization allows us to obtain efficient ANN models for classification problems.