Abstract
In the report a non-classical type of neural network is described. The structure of neural network (NN) takes into consideration both differences among inputs of neural network and “physical nature” of pattern recognition process. In this respect a different combinations of various types of inputs of neural network are singled out in advance. Method of the development based on a special fuzzy operator and a problem of correctness network's solution is being considered as problem of solvability for operator equation. The construction gives opportunity to state a problem of existence and finding exact solution for pattern recognition problem.