Neuronet models for complex process description

A methodology for developing a model of a complex multistage process based on multi-layer neuron network is described. The neuronet model structure for a multi-stage process and an algorithm for its formation are described. The neuronet model learning procedure is considered. The paper shows that the learning process can be reduced to the minimization of a multivariable function. The equations for analytical recalculation of loss function's gradient are derived that allow the application of effective optimization techniques for network learning.