Qualia & Spleat, an end to end neuromorphic workflow

Description

Spiking neural networks (SNNs) have shown a remarkable 5 to 8-fold increase in efficiency¹ when compared to formal neural networks(FNNs), particularly in anticipation of an implementation on a dedicated hardware accelerator. In order to obtain this gain in practice, it'snecessary to master the training, quantization, and deployment steps dédicated to a specific hardware. By the combination of ourFramework QUALIA and our SNN configurable architecture SPLEAT, we propose an end to end neuromorphic workflow.

Abstract

International audience

Additional details