А.С. Makarenko, V.M. Bilyuga

In the given work the model of neural network which has the multiple-valued solutions is explored. The basic feature is that in the activation function of neuron anticipation is entered. Thus, if on the entrance of neural network to give some entrance vector, on an output not single value can be received, as in the case of the already explored models. That is the feature of model that a multivaluedness appears in the case of sigmoid - like activation function, which is intrinsic to usual artificial neural models. For simplification of program realization as the basic case the neural Hopfield network with the piece-linear activation function had been chosen, that also the conduct of model enables analytically to explore. The possible consequences of the received solutions for the principles of operation of the biological systems (in particular functioning of consciousness) are considered, and also consequences for establishing of work principles of new types of the computer systems which would have the inherit principles of work common to the living systems are discussed.