Integrated neuromorphic computing networks by artificial spin synapses and spin neurons

Abstract

One long-standing goal in the emerging neuromorphic field is to create a reliable neural network hardware implementation that has low energy consumption, while providing massively parallel computation. Although diverse oxide-based devices have made significant progress as artificial synaptic and neuronal components, these devices still need further optimization regarding linearity, symmetry, and stability. Here, we present a proof-of-concept experiment for integrated neuromorphic computing networks by utilizing spintronics-based synapse (spin-S) and neuron (spin-N) devices, along with linear and symmetric weight responses for spin-S using a stripe domain and activation functions for spin-N. An integrated neural network of electrically connected spin-S and spin-N successfully proves the integration function for a simple pattern classification task. We simulate a spin-N network using the extracted device characteristics and demonstrate a high classification accuracy (over 93%) for the spin-S and spin-N optimization without the assistance of additional software or circuits required in previous reports. These experimental studies provide a new path toward establishing more compact and efficient neural network systems with optimized multifunctional spintronic devices.

Compensated Ferrimagnet Based Artificial Synapse and Neuron for Ultrafast Neuromorphic Computing

“…The simulated ANN based on the artificial synapses and neurons could perform the digital recognition tasks with an accuracy rate of over 93%. The writing efficiency of the neuromorphic computing based on FIM is 10 6 times higher than that of FM-based neuromorphic computing, [48][49][50][51] and 10 times higher than that of the state-of-art FM/AFM heterojunction-based neuromorphic computing. [54] Our results thus illustrate the potential of FIM in realizing high-efficiency spin artificial neuromorphic computing.…”

“…[46,47] Experimentally, the SOT-induced magnetization switching and the DW motion in ferromagnets (FMs) have been implemented for mimicking synapses and neurons. [48][49][50][51] These early results were accomplished, however, by using relatively long writing pulses with durations in millisecond time scale. For realizing a faster neuromorphic response, one could explore the fast spin dynamics in antiferromagnets (AFMs), [52] or compensated ferrimagnets (FIMs).…”

Quantum Artificial Synapses

“…In addition, a proof-of-principle neuromorphic computing networks using MgO/Co 20 Fe 60 B 20 /W spintronic devices were reported by Yang et al most lately. [99] The integrated neural networks that consist of spintronics-based synapses and neurons were used to perform the typical pattern classification task, where an excellent classification accuracy over 93% was obtained. It would be meaningful to establish more compact and efficient spintronics-based neural network systems to realize diverse cognition functions.…”

“…Reproduced under the terms of a Creative Commons Attribution 4.0 License (CC BY). [99] Copyright 2021, The Authors, published by Springer Nature. e) Synapse-inspired hotelectron tunnel junction memristor showing programmable multi-level programming electrical and optical properties.…”

“…d) Integrated multi-layer neural networks consisting of spintronics-based synapses and neurons. Reproduced under the terms of a Creative Commons Attribution 4.0 License (CC BY) [99]. Copyright 2021, The Authors, published by Springer Nature.…”