A simple PSTDP circuit for Analog Implementation of Spiking Neural Networks

Abstract

Analog implementations of spiking neural networks allow stimuli from real-life situations to be applied directly to such systems resulting in efficient neural processing architectures. Such systems are capable of adapting to the environment due to the capacity to learn. One of the learning rules known as Spike Time Dependent Plasticity (STDP) modifies the efficacy of the synapse based on the time difference between the pre and post synaptic pulses. This paper presents a simple and novel analog implementation of pairbased STDP circuit that exhibits both potentiation and depression. The proposed PSTDP circuit is used to build a simple spiking neural network along with integrate and fire neuron and differential pair synapse circuits that are available in literature. The network operation is demonstrated by applying varying pulses as pre and post spikes to the PSTDP circuit and observing the change in output of the postsynaptic neuron circuit as the synaptic weights are getting modified. The circuit is implemented in Cadence Virtuoso with GPDK 180 nm technology parameters.