Competitive Maximization of Neuronal Activity in Convolutional Recurrent Spiking Neural Networks

Abstract

Spiking neural networks (SNNs) are the promising algorithm for specific neurochip hardware real-time solutions. SNNs are believed to be highly energy and computationally efficient. We focus on developing local learning rules that are capable to provide both supervised and unsupervised learning. We suppose that each neuron in a biological neural network tends to maximize its activity in competition with other neurons. This principle was put at the basis of SNN learning algorithm called FEELING. Here we introduce efficient Convolutional Recurrent Spiking Neural Network architecture that uses FEELING rules and provides better results than fully connected SNN on MNIST benchmark having 55 times less learnable weight parameters.