Asymmetric Synaptic Plasticity Based on Arbitrary Pre- and Postsynaptic Timing Spikes Using Finite State Model

A new computational synaptic plasticity model is presented to embody the relative firing rates between pre-and postsynaptic spikes. The proposed synaptic plasticity model is developed in two steps. Firstly, a finite state model is used to explain diverse protocols of interactive firing spikes, in which an action is produced over state transition to change synaptic efficacy. The produced action denotes synaptic efficacy change rate reliant on a double-stochastic process. Secondly, the total synaptic efficacy update is defined as a nonlinear bounded function dependent on both synaptic efficacy change rate and previous synaptic efficacy value. The proposed synaptic efficacy model is tested experimentally and shows a high degree of similarities with biological data. The inherently determined critical timing window for pre-and-postsynaptic spike pair has a non-zero time shift for peak potentiation or depression, which agrees with the biological synaptic transmission delays in axons and dendrites. Further, numerical analysis shows that the present model is in agreement with well-accepted synaptic learning rule, BCM.