SONIC: A Sparse Neural Network Inference Accelerator with Silicon Photonics for Energy-Efficient Deep Learning

TL;DR

SONIC is a silicon photonics-based accelerator that leverages sparsity in neural networks to significantly improve energy efficiency and performance, outperforming existing electronic and photonic accelerators.

Contribution

This paper introduces SONIC, a novel silicon photonics-based accelerator that co-designs hardware and software to exploit neural network sparsity for energy-efficient inference.

Findings

Up to 5.8x better performance-per-watt than electronic accelerators

Up to 8.4x lower energy-per-bit than electronic accelerators

Up to 13.8x better performance-per-watt than photonic accelerators

Abstract

Sparse neural networks can greatly facilitate the deployment of neural networks on resource-constrained platforms as they offer compact model sizes while retaining inference accuracy. Because of the sparsity in parameter matrices, sparse neural networks can, in principle, be exploited in accelerator architectures for improved energy-efficiency and latency. However, to realize these improvements in practice, there is a need to explore sparsity-aware hardware-software co-design. In this paper, we propose a novel silicon photonics-based sparse neural network inference accelerator called SONIC. Our experimental analysis shows that SONIC can achieve up to 5.8x better performance-per-watt and 8.4x lower energy-per-bit than state-of-the-art sparse electronic neural network accelerators; and up to 13.8x better performance-per-watt and 27.6x lower energy-per-bit than the best known photonic neural network accelerators.