Effect of random synaptic dilution in oscillator neural networks

Abstract

To discuss robustness against damage of synaptic connections in oscillator neural networks, we examine the effect of random dilution of the synapses in the case in which connections are symmetric. Deriving the mean-field equations in the replica symmetry approximation, we obtain the storage capacity and the critical overlap as functions of the ratio of the disconnected synapses. In addition, we find that theoretical results for this system are supported by our numerical simulations. Finally, it is suggested that the oscillator neural network is more robust against dilution than is the Hopfield model.