Scholarlink

<p>Neuromorphic computing could be used to create artificial intelligence with high compactness and efficiency. However, complementary metal–oxide–semiconductor (CMOS) circuits are inherently different to biological neurons, and intricate CMOS circuits are needed to realize neuromorphic behaviours. Diffusive memristors are based on ion dynamics and have similarities with biological neurons. They could, thus, be used to create energy- and area-efficient neuromorphic systems. Here we describe a spiking artificial neuron comprising one diffusive memristor, one transistor and one resistor (1M1T1R), which occupies the footprint of a single transistor when vertically integrated. Our neuron exhibits six key neuronal characteristics: leaky integration, threshold firing, cascaded connection, intrinsic plasticity, refractory period and stochasticity. The energy consumption of our 1M1T1R neuron reaches the picojoule per spike level and could reach attojoule per spike levels with further scaling. We simulate a recurrent spiking neural network based on our artificial neuron model and show the impact of the key neuronal characteristics on system performance.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

A spiking artificial neuron based on one diffusive memristor, one transistor and one resistor

  • 期刊论文

  • 发表日期：

    2025年10月27日

查看全文
  • Ruoyu Zhao,
  • Tong Wang,
  • Taehwan Moon,
  • Yichun Xu,
  • Jian Zhao,
  • Piyush Sud,
  • Seung Ju Kim,
  • Han-Ting Liao,
  • Ye Zhuo,
  • Rivu Midya,
  • Shiva Asapu,
  • Dawei Gao,
  • Zixuan Rong,
  • Qinru Qiu,
  • Cynthia Bowers,
  • Krishnamurthy Mahalingam,
  • S. Ganguli,
  • A. K. Roy,
  • Qing Wu,
  • Jin-Woo Han,
  • R. Stanley Williams,
  • Yong Chen,
  • J. Joshua Yang

摘要

Neuromorphic computing could be used to create artificial intelligence with high compactness and efficiency. However, complementary metal–oxide–semiconductor (CMOS) circuits are inherently different to biological neurons, and intricate CMOS circuits are needed to realize neuromorphic behaviours. Diffusive memristors are based on ion dynamics and have similarities with biological neurons. They could, thus, be used to create energy- and area-efficient neuromorphic systems. Here we describe a spiking artificial neuron comprising one diffusive memristor, one transistor and one resistor (1M1T1R), which occupies the footprint of a single transistor when vertically integrated. Our neuron exhibits six key neuronal characteristics: leaky integration, threshold firing, cascaded connection, intrinsic plasticity, refractory period and stochasticity. The energy consumption of our 1M1T1R neuron reaches the picojoule per spike level and could reach attojoule per spike levels with further scaling. We simulate a recurrent spiking neural network based on our artificial neuron model and show the impact of the key neuronal characteristics on system performance.