Building Neural Network for Pattern Recognition

The paper considers the problem of pattern recognition using neural networks exemplified by patients' bone fracture images. The conventional approach has been used to build a deep-belief neural network, in which the input, processing, and output neuron layers are formed. The training of neural networks is based on the use of a multilayer artificial neural network (hereinafter referred to as ANN). Interneuronal connections of a multilayer network are tuned and the weighting coefficients calculated using the algorithm developed. This algorithm allows building the best ANN architecture. This algorithm reliably allows completing the ANN training. Processing the training set has given an almost zero recognition error, which indicates the advantages of this architecture forming technique over others.