Neural network model for control of complicated systems

The article describes the principle of forming a classifier based on multilayer neural networks, allowing to solve tasks of control of technical condition of systems of high dimensionality, when little of a priori information about the current technical condition, there exist measurement error of parameters and uncertainty of environmental factors. Developed the model of control of technical condition with an excess number of non-linear converters in the layer. To increase the generalizing ability of the model of control used original rules output error.