Enhanced Memory Association on Neural Network with Dynamic Synapses

We investigate influences of transitive dynamics induced by dynamic synapses on a network model that performs a memory association task, which require to recall memory patterns from the partially disturbed and time-varying sensory input. The model is based on an associative memory network with dynamic synapses whose connection strength changes with short-term plasticity (STP) mechanism. We evaluate the memory association performances on model parameters and compare dynamic and static synapses by numerical simulation. The performance depends on the balance of strength of feedforward and recurrent synaptic connections, internal noise, and the properties of dynamic synapses. Further, the network with dynamic synapses predominates over the network with static synapses.