Synthesis method of robust neural network models of systems and processes

Abstract

However, we have mathematical problems which hamper a wide application of these models for solving practical tasks. The first problem is that neural model formation problems are treated as ill-conditioned. However, these problems can be considered as conditionally correct of structural parameter optimization, when model structure is known, but its parameters are not known. In this case, to find learning parameters (ANN), the machine learning methods relying on the use of decision synthesis regulating algorithms, including deep learning methods, are widely used.