An Energy-Efficient Mixed-Signal Neuron for Inherently Error-Resilient Neuromorphic Systems

TL;DR

This paper introduces a mixed-signal neuron design that significantly improves energy efficiency for neuromorphic systems, leveraging CNN error resilience to tolerate inherent circuit noise, demonstrated on MNIST and CIFAR-10 datasets.

Contribution

The paper presents a novel mixed-signal neuron design with superior energy efficiency and analyzes its error resilience in CNNs, validated through system-level simulations.

Findings

2-3 orders of magnitude energy efficiency improvement over digital neurons

CNN error resilience allows handling of circuit noise without accuracy loss

System-level analysis shows only 3% increase in error with noise power ~1 μV^2

Abstract

This work presents the design and analysis of a mixed-signal neuron (MS-N) for convolutional neural networks (CNN) and compares its performance with a digital neuron (Dig-N) in terms of operating frequency, power and noise. The circuit-level implementation of the MS-N in 65 nm CMOS technology exhibits 2-3 orders of magnitude better energy-efficiency over Dig-N for neuromorphic computing applications - especially at low frequencies due to the high leakage currents from many transistors in Dig-N. The inherent error-resiliency of CNN is exploited to handle the thermal and flicker noise of MS-N. A system-level analysis using a cohesive circuit-algorithmic framework on MNIST and CIFAR-10 datasets demonstrate an increase of 3% in worst-case classification error for MNIST when the integrated noise power in the bandwidth is ~ 1 {\mu}V2.