Some approaches of improving the quality of artificial neural network training

Abstract

The paper is devoted to a problem of regular improving the quality of artificial neural network (ANN) training. The object of study is a complex neural network which consists of 2-dimensional Kohonen network and Wilshaw and von der Malsburg network. These networks are applied to a timetable problem for transport systems. The main existing results of using optimal control theory for ANN training are analyzed; authors suggest a new technique based on the direct neural control. Authors give comparative values of error during a training process for the traditional methods and a new approach. It is presented that this new technique is better than the traditional one for considered neural networks.