Аннотация:In artificial neural networks the simplest attractor types of dynamic systems are used for information coding (as a rule, stable focuses). System motion from some initial point of phase space to a stationary state corresponds to a process of image recognition. Informative capacity of a neural network is governed by topology of the global system attractor (in the simplest case by number of possible stationary system states). In the work properties of the global attractor of one of the simplest models of a neuron-like optical system with nonlocal couplings are studied.