EqSpike: Spike-driven Equilibrium Propagation for Neuromorphic Implementations

TL;DR

EqSpike introduces a spike-driven learning algorithm based on Equilibrium Propagation for neuromorphic systems, achieving high accuracy and significant energy efficiency, with biological plausibility in weight updates.

Contribution

The paper develops EqSpike, a novel spike-based learning algorithm compatible with neuromorphic hardware, demonstrating high accuracy and energy efficiency.

Findings

97.6% accuracy on MNIST

Reduces inference energy by 1000x compared to GPUs

Weight updates resemble Spike Timing Dependent Plasticity

Abstract

Finding spike-based learning algorithms that can be implemented within the local constraints of neuromorphic systems, while achieving high accuracy, remains a formidable challenge. Equilibrium Propagation is a promising alternative to backpropagation as it only involves local computations, but hardware-oriented studies have so far focused on rate-based networks. In this work, we develop a spiking neural network algorithm called EqSpike, compatible with neuromorphic systems, which learns by Equilibrium Propagation. Through simulations, we obtain a test recognition accuracy of 97.6% on MNIST, similar to rate-based Equilibrium Propagation, and comparing favourably to alternative learning techniques for spiking neural networks. We show that EqSpike implemented in silicon neuromorphic technology could reduce the energy consumption of inference and training respectively by three orders and two orders of magnitude compared to GPUs. Finally, we also show that during learning, EqSpike weight updates exhibit a form of Spike Timing Dependent Plasticity, highlighting a possible connection with biology.