On the Role of Synaptic Depression in the Performance of Attractor Neural Networks

Using a biologically motivated model of synaptic depression and within a mean-field approach, we examined the role of synaptic depression in the capacity of a binary neural network with N units to store and retrieve P patterns. In the limit of α ≡P/N → 0, our results demonstrate the appearance of a novel phase characterized by quick transitions from one memory state to another. This phenomenon might reflect the flexibility of real neural systems to receive and respond to novel and changing external stimuli. In addition, we have computed the maximum storage capacity of such a network in the limit of α = 0 and T = 0. Supported by mean-field results and Monte Carlo simulations, we concluded that the critical storage capacity for effective retrieval of stable memory patterns decreases with the degree of the depression. Nevertheless, the storage of memories as oscillatory states will require a different definition of storage capacity. How such a new storage capacity depends on the synaptic depression is still an open question.