EVALUATION OF THE PROPERTIES OF NEURAL NETWORK MODELS IN PROBLEMS OF DIAGNOSIS AND AUTOMATIC CLASSIFICATION.
The complex of indicators for measuring the properties of neural networks is proposed. The application of the proposed indicators in practice allows to automate the formation process of neural network models for solving problems of diagnostics and automatic classification problems.
Keywords: neural network, plasticity, homogeneity, parallelism, hierarchy, preorganization, functional modularity, generalization, learning, interpretability, reliability, memory allocation, variability, efficiency
