A Study on the Memory of Temporal Sequences Using a Pulse-type Hardware Neural Network with STDP

摘要

Synaptic plasticity, which is dependent on the order of and interval between pre- and postsynaptic spikes (STDP: spike-timing-dependent synaptic plasticity), has recently been observed by physiological experiments. There are two types of STDP, one characterized by an asymmetric time window and the other by a symmetric time window (a Mexican hat type window). The literature reports that a Mexican-hat window participates in an inhibitory interneuron. In this study, we investigate the memory of temporal sequences using a pulse-type hardware neural network (P-HNN) with STDP (a Mexican-hat time window). Specifically, we propose the construction of a P-HNN with STDP. In addition, we examine the temporal sequence patterns which were stored by using a P-HNN with STDP. As a result, we show that a P-HNN with STDP stores temporal sequence output voltage patterns in a way that conforms to the temporal sequence input current patterns.