SupeRBNN: Randomized Binary Neural Network Using Adiabatic Superconductor Josephson Devices

TL;DR

SupeRBNN introduces an AQFP-based randomized binary neural network acceleration framework that significantly improves energy efficiency through software-hardware co-optimization and innovative circuit design, enabling practical superconducting BNN accelerators.

Contribution

This work presents the first comprehensive AQFP-based BNN acceleration framework with novel stochastic accumulation and circuit optimization techniques.

Findings

Achieves 78,000x higher energy efficiency than ReRAM-based BNNs.

Maintains comparable model accuracy with significantly improved energy efficiency.

Demonstrates at least 100x higher energy efficiency compared to other superconducting approaches.

Abstract

Adiabatic Quantum-Flux-Parametron (AQFP) is a superconducting logic with extremely high energy efficiency. By employing the distinct polarity of current to denote logic `0' and `1', AQFP devices serve as excellent carriers for binary neural network (BNN) computations. Although recent research has made initial strides toward developing an AQFP-based BNN accelerator, several critical challenges remain, preventing the design from being a comprehensive solution. In this paper, we propose SupeRBNN, an AQFP-based randomized BNN acceleration framework that leverages software-hardware co-optimization to eventually make the AQFP devices a feasible solution for BNN acceleration. Specifically, we investigate the randomized behavior of the AQFP devices and analyze the impact of crossbar size on current attenuation, subsequently formulating the current amplitude into the values suitable for use in BNN computation. To tackle the accumulation problem and improve overall hardware performance, we propose a stochastic computing-based accumulation module and a clocking scheme adjustment-based circuit optimization method. We validate our SupeRBNN framework across various datasets and network architectures, comparing it with implementations based on different technologies, including CMOS, ReRAM, and superconducting RSFQ/ERSFQ. Experimental results demonstrate that our design achieves an energy efficiency of approximately 7.8x10^4 times higher than that of the ReRAM-based BNN framework while maintaining a similar level of model accuracy. Furthermore, when compared with superconductor-based counterparts, our framework demonstrates at least two orders of magnitude higher energy efficiency.