Multilayer Magnetic Domain Wall MTJ-based Spiking Neural Network

Aijaz H. Lone, Daniel N. Rahimi, Hossein Fariborzi, Gianluca Setti

Research output: Chapter in Book/Report/Conference proceedingConference contributionpeer-review

Abstract

Spintronic devices, especially the magnetic tunnel junction and magnetic domain wall-based devices, hold significant promise for applications in energy-efficient data storage and Unconventional computing architectures. We present a novel multilayer spintronic neuromorphic device based on spin-orbit torque-driven domain wall dynamics. The typical leaky integrate and fire LIF neuron-like characteristics are realized using the combination of SOT and demagnetization energy effects. The device characteristics are modelled as the modified LIF model. We test the spiking neuron model for the classification of the MNIST dataset by emulating a 3-layer spiking neural network SNN -based on the DW-MTJ LIF neuron model. The network achieves classification accuracy above 96% thus the proposed device can be integrated with the CMOS for energy efficient neuromorphic computing.

Original languageEnglish (US)
Title of host publication2024 IEEE 24th International Conference on Nanotechnology, NANO 2024
PublisherIEEE Computer Society
Pages146-149
Number of pages4
ISBN (Electronic)9798350386240
DOIs
StatePublished - 2024
Event24th IEEE International Conference on Nanotechnology, NANO 2024 - Gijon, Spain
Duration: Jul 8 2024Jul 11 2024

Publication series

NameProceedings of the IEEE Conference on Nanotechnology
ISSN (Print)1944-9399
ISSN (Electronic)1944-9380

Conference

Conference24th IEEE International Conference on Nanotechnology, NANO 2024
Country/TerritorySpain
CityGijon
Period07/8/2407/11/24

Keywords

  • Domain wall devices
  • Leaky-integrate and fire neurons
  • Magnetic tunnel junction
  • Neuromorphic Computing
  • Spiking neural networks (SNN)
  • Spintronics

ASJC Scopus subject areas

  • Bioengineering
  • Electrical and Electronic Engineering
  • Materials Chemistry
  • Condensed Matter Physics

Fingerprint

Dive into the research topics of 'Multilayer Magnetic Domain Wall MTJ-based Spiking Neural Network'. Together they form a unique fingerprint.

Cite this