Progressive Compression and Weight Reinforcement for Spiking Neural Networks

Compression progressive et renforcement du poids pour les réseaux de neurones a impulsions

Hammouda Elbez , Kamel Benhaoua , Philippe Devienne (1) , Pierre Boulet (1)

1 CRIStAL - Centre de Recherche en Informatique, Signal et Automatique de Lille - UMR 9189

Neuromorphic architectures are one of the most promising architectures to significantly reduce the energy consumption of tomorrow’s computers. These architectures are inspired by the behaviour of the brain at a fairly precise level and consist of artificial Spiking Neural Networks (SNNs). To optimise the implementation of these architectures, we propose in this paper a novel progressive network compression and reinforcement technique based on two functions, progressive pruning and dynamic synaptic weight reinforcement used after each training batch. The proposed approach delivers a highly compressed network (up to 75 % of compression rate) while preserving the network performance when tested with MNIST.

Domaines: Intelligence artificielle [cs.AI]; Apprentissage [cs.LG]

Hammouda Elbez, Kamel Benhaoua, Philippe Devienne, Pierre Boulet. Progressive Compression and Weight Reinforcement for Spiking Neural Networks. Concurrency and Computation: Practice and Experience, 2022, ⟨10.1002/cpe.6891⟩. ⟨hal-02737057⟩