Abstract: | A stochastic memory model based on the net of excitatory and inhibitory neurons is considered. Its learning ability is due to a decrease of the reactivity of inhibitory elements. Information capacity of such neuron net was estimated in relation to its structure parameters. It has been shown that in the definite range of their change the information capacity of the net under consideration significantly exceeds that of the neuron net with the memory based on the plasticity of excitatory elements. |