SNE: an Energy-Proportional Digital Accelerator for Sparse Event-Based Convolutions

TL;DR

This paper introduces a digital accelerator optimized for sparse event-based neural network processing, achieving high energy efficiency and proportional computation for event streams from sensors.

Contribution

The paper presents a novel energy-proportional digital accelerator for sparse event-based CNNs, with the lowest reported energy per operation on neuromorphic hardware.

Findings

Achieves 4.5 TOP/s/W energy efficiency

Operates with 4-bit quantized event-based CNNs

Lowest energy per operation reported on digital neuromorphic engines

Abstract

Event-based sensors are drawing increasing attention due to their high temporal resolution, low power consumption, and low bandwidth. To efficiently extract semantically meaningful information from sparse data streams produced by such sensors, we present a 4.5TOP/s/W digital accelerator capable of performing 4-bits-quantized event-based convolutional neural networks (eCNN). Compared to standard convolutional engines, our accelerator performs a number of operations proportional to the number of events contained into the input data stream, ultimately achieving a high energy-to-information processing proportionality. On the IBM-DVS-Gesture dataset, we report 80uJ/inf to 261uJ/inf, respectively, when the input activity is 1.2% and 4.9%. Our accelerator consumes 0.221pJ/SOP, to the best of our knowledge it is the lowest energy/OP reported on a digital neuromorphic engine.