The Realisation of Neural Network Structural Optimization Algorithm

Abstract

Abstract-This paper presents a deep analysis of literature on the problems of optimization of parameters and structure of the neural networks and the basic disadvantages that are present in the observed algorithms and methods. As a result, there is suggested a new algorithm for neural network structure optimization, which is free of the major shortcomings of other algorithms. The paper describes a detailed description of the algorithm, its implementation and application for recognition problems.

Neural Network Structure Optimization Algorithm

“…Internally neural networks are presented as numeric matrix sequences of each layer weight except for the input one [1]. In Fig.1 the matrix sequence for [2-3-2] network type is shown: hidden layer matrix 2x3 and output layer one 3x2.…”

“…Human Recognition. The implemented program system is used to research problems of human face recognition [1]. The face image database of Yale university was used as output data [32].…”

“…The unit of neural networks is widely used to solve various problems including recognition tasks. The existence of a method for automatic search of neural network optimal structure could provide an opportunity to get the structure of a neural network much faster, that would better suit the subject area and existing incoming data [1].…”

“…The algorithm described in this article addresses all of these issues. The main goal of this work is to analyze the structure optimization algorithm of neural network during its learning for the tasks of pattern recognition [24] and to implement the algorithm using program instruments [1].…”