Topology optimization of random memristors for input-aware dynamic SNN.

Journal: Science advances
PMID:

Abstract

Machine learning has advanced unprecedentedly, exemplified by GPT-4 and SORA. However, they cannot parallel human brains in efficiency and adaptability due to differences in signal representation, optimization, runtime reconfigurability, and hardware architecture. To address these challenges, we introduce pruning optimization for input-aware dynamic memristive spiking neural network (PRIME). PRIME uses spiking neurons to emulate brain's spiking mechanisms and optimizes the topology of random memristive SNNs inspired by structural plasticity, effectively mitigating memristor programming stochasticity. It also uses the input-aware early-stop policy to reduce latency and leverages memristive in-memory computing to mitigate von Neumann bottleneck. Validated on a 40-nm, 256-K memristor-based macro, PRIME achieves comparable classification accuracy and inception score to software baselines, with energy efficiency improvements of 37.8× and 62.5×. In addition, it reduces computational loads by 77 and 12.5% with minimal performance degradation and demonstrates robustness to stochastic memristor noise. PRIME paves the way for brain-inspired neuromorphic computing.

Authors

  • Bo Wang
    Department of Clinical Laboratory Medicine Center, Inner Mongolia Autonomous Region People's Hospital, Hohhot, Inner Mongolia, China.
  • Xinyuan Zhang
    Department of Electrical and Electronic Engineering, The University of Hong Kong, Hong Kong, China.
  • Shaocong Wang
    Department of Electrical and Electronic Engineering, The University of Hong Kong, Hong Kong, China.
  • Ning Lin
    Department of Electrical and Electronic Engineering, The University of Hong Kong, Hong Kong, China.
  • Yi Li
    Wuhan Zoncare Bio-Medical Electronics Co., Ltd, Wuhan, China.
  • Yifei Yu
    Department of Electrical and Electronic Engineering, The University of Hong Kong, Hong Kong, China.
  • Yue Zhang
    Department of Ophthalmology, Beijing Hospital, National Center of Gerontology, Institute of Geriatric Medicine, Chinese Academy of Medical Sciences, Beijing, China.
  • Jichang Yang
    Department of Electrical and Electronic Engineering, The University of Hong Kong, Hong Kong, China.
  • Xiaoshan Wu
    Department of Anesthesiology, the First Affiliated Hospital of Guangxi Medical University, Nanning, Guangxi, People's Republic of China.
  • Yangu He
    Department of Electrical and Electronic Engineering, The University of Hong Kong, Hong Kong, China.
  • Songqi Wang
    Department of Electrical and Electronic Engineering, The University of Hong Kong, Hong Kong, China.
  • Tao Wan
    Department of Electrical and Electronic Engineering, The University of Hong Kong, Hong Kong, China.
  • Rui Chen
    College of Food Science and Engineering, Northwest A&F University, Yangling 712100, Shanxi, China.
  • Guoqi Li
    University of Chinese Academy of Sciences, Beijing 100049, China.
  • Yue Deng
    School of Artificial Intelligence, Beihang University, Beijing 100191, China.
  • Xiaojuan Qi
    Department of Electrical and Electronic Engineering, The University of Hong Kong, Hong Kong, China.
  • Zhongrui Wang
    Department of Electrical and Electronic Engineering, The University of Hong Kong, Hong Kong, China.
  • Dashan Shang
    State Key Lab of Fabrication Technologies for Integrated Circuits, Institute of Microelectronics, Chinese Academy of Sciences, Beijing 100029, China.