Domain wall motion-based dual-threshold activation unit for low-power classification of non-linearly separable functions
Recently, a great deal of scientific endeavour has been devoted to developing spin-based neuromorphic platforms owing to the ultra-low-power benefits offered by spin devices and the inherent correspondence between spintronic phenomena and the desired neuronal, synaptic behavior. While domain wall mo...
Saved in:
Main Authors: | , , , , |
---|---|
Other Authors: | |
Format: | Article |
Language: | English |
Published: |
2020
|
Subjects: | |
Online Access: | https://hdl.handle.net/10356/141591 |
Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
Institution: | Nanyang Technological University |
Language: | English |
id |
sg-ntu-dr.10356-141591 |
---|---|
record_format |
dspace |
spelling |
sg-ntu-dr.10356-1415912020-06-09T06:30:58Z Domain wall motion-based dual-threshold activation unit for low-power classification of non-linearly separable functions Deb, Suman Vatwani, Tarun Chattopadhyay, Anupam Basu, Arindam Fong, Xuanyao School of Computer Science and Engineering School of Electrical and Electronic Engineering Engineering::Computer science and engineering ANN Domain Wall Motion Recently, a great deal of scientific endeavour has been devoted to developing spin-based neuromorphic platforms owing to the ultra-low-power benefits offered by spin devices and the inherent correspondence between spintronic phenomena and the desired neuronal, synaptic behavior. While domain wall motion-based threshold activation unit has previously been demonstrated for neuromorphic circuits, it remains well known that neurons with threshold activation cannot completely learn nonlinearly separable functions. This paper addresses this fundamental limitation by proposing a novel domain wall motion-based dual-threshold activation unit with additional nonlinearity in its function. Furthermore, a new learning algorithm is formulated for a neuron with this activation function. We perform 100 trials of tenfold training and testing of our neural networks on real-world datasets taken from the UCI machine learning repository. On an average, the proposed algorithm achieves 1.04 × -6.54 × lower misclassification rate (MCR) than the traditional perceptron learning algorithm. In a circuit-level simulation, the neural networks with the proposed activation unit are observed to outperform the perceptron networks by as much as 2.98 × MCR. The energy consumption of a neuron having the proposed domain wall motion-based activation unit averages to 35 fJ approximately. 2020-06-09T06:30:58Z 2020-06-09T06:30:58Z 2018 Journal Article Deb, S., Vatwani, T., Chattopadhyay, A., Basu, A., & Fong, X. (2018). Domain wall motion-based dual-threshold activation unit for low-power classification of non-linearly separable functions. IEEE Transactions on Biomedical Circuits and Systems, 12(6), 1410-1421. doi:10.1109/TBCAS.2018.2867038 1932-4545 https://hdl.handle.net/10356/141591 10.1109/TBCAS.2018.2867038 30176604 2-s2.0-85052663152 6 12 1410 1421 en IEEE Transactions on Biomedical Circuits and Systems © 2018 IEEE. All rights reserved. |
institution |
Nanyang Technological University |
building |
NTU Library |
country |
Singapore |
collection |
DR-NTU |
language |
English |
topic |
Engineering::Computer science and engineering ANN Domain Wall Motion |
spellingShingle |
Engineering::Computer science and engineering ANN Domain Wall Motion Deb, Suman Vatwani, Tarun Chattopadhyay, Anupam Basu, Arindam Fong, Xuanyao Domain wall motion-based dual-threshold activation unit for low-power classification of non-linearly separable functions |
description |
Recently, a great deal of scientific endeavour has been devoted to developing spin-based neuromorphic platforms owing to the ultra-low-power benefits offered by spin devices and the inherent correspondence between spintronic phenomena and the desired neuronal, synaptic behavior. While domain wall motion-based threshold activation unit has previously been demonstrated for neuromorphic circuits, it remains well known that neurons with threshold activation cannot completely learn nonlinearly separable functions. This paper addresses this fundamental limitation by proposing a novel domain wall motion-based dual-threshold activation unit with additional nonlinearity in its function. Furthermore, a new learning algorithm is formulated for a neuron with this activation function. We perform 100 trials of tenfold training and testing of our neural networks on real-world datasets taken from the UCI machine learning repository. On an average, the proposed algorithm achieves 1.04 × -6.54 × lower misclassification rate (MCR) than the traditional perceptron learning algorithm. In a circuit-level simulation, the neural networks with the proposed activation unit are observed to outperform the perceptron networks by as much as 2.98 × MCR. The energy consumption of a neuron having the proposed domain wall motion-based activation unit averages to 35 fJ approximately. |
author2 |
School of Computer Science and Engineering |
author_facet |
School of Computer Science and Engineering Deb, Suman Vatwani, Tarun Chattopadhyay, Anupam Basu, Arindam Fong, Xuanyao |
format |
Article |
author |
Deb, Suman Vatwani, Tarun Chattopadhyay, Anupam Basu, Arindam Fong, Xuanyao |
author_sort |
Deb, Suman |
title |
Domain wall motion-based dual-threshold activation unit for low-power classification of non-linearly separable functions |
title_short |
Domain wall motion-based dual-threshold activation unit for low-power classification of non-linearly separable functions |
title_full |
Domain wall motion-based dual-threshold activation unit for low-power classification of non-linearly separable functions |
title_fullStr |
Domain wall motion-based dual-threshold activation unit for low-power classification of non-linearly separable functions |
title_full_unstemmed |
Domain wall motion-based dual-threshold activation unit for low-power classification of non-linearly separable functions |
title_sort |
domain wall motion-based dual-threshold activation unit for low-power classification of non-linearly separable functions |
publishDate |
2020 |
url |
https://hdl.handle.net/10356/141591 |
_version_ |
1681057260503040000 |