Neural network analysis for image classification

Abstract

The article considers the possibility of modeling artificial neural networks using the mathematical apparatus of information theory. The issues of pattern recognition, classification and clustering of images using neural networks are represented by two main architectures: a direct distribution network and convolutional networks. The possibility of using orthogonal transformations to increase the efficiency of neural networks, the use of wavelet transformations in convolutional networks is investigated. Based on the theoretical studies carried out, the directions on practical application of the obtained results are proposed.