Feature Extraction using Spiking Convolutional Neural Networks

Spiking neural networks are biologically plausible counterparts of the artificial neural networks. Conventional (non spiking) artificial neural networks are trained using a stochastic gradient descent algorithm (back propagation) while spiking neural networks are trained using the spike timing dependant plasticity algorithm. Training conventional neural networks is a memory and computationally/energy intensive job. Spiking networks show the potential to reduce both memory and energy usage. In this work, we consider spiking neural networks (SNN) as a classifier of the MNIST and NM-NIST digits. We compare SNNs with convolutional neural networks in terms of classification accuracy, catastrophic forgetting.