Biologically Inspired Temporal Sequence Learning

We propose a temporal sequence learning model in spiking neural networks consisting of Izhikevich spiking neurons. In our reward-based learning model, we train a network to associate two stimuli with temporal delay and a target response. Learning rule is dependent on reward signals that modulate the weight changes derived from spike-timing dependent plasticity (STDP) function. The dynamic properties of our model can be attributed to the sparse and recurrent connectivity, synaptic transmission delays, background activity and inter-stimulus interval (ISI). We have tested the learning in visual recognition task, and temporal AND and XOR problems. The network can be trained to associate a stimulus pair with its target response and to discriminate the temporal sequence of the stimulus presentation.