Coupled VO2 Oscillators Circuit as Analog First Layer Filter in Convolutional Neural Networks
In this work we present an in-memory computing platform based on coupled VO2 oscillators fabricated in a crossbar configuration on silicon. Compared to existing platforms, the crossbar configuration promises significant improvements in terms of area density and oscillation frequency. Further, the cr...
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Frontiers Media S.A.
2021-02-01
|
| Series: | Frontiers in Neuroscience |
| Subjects: | |
| Online Access: | https://www.frontiersin.org/articles/10.3389/fnins.2021.628254/full |
| id |
doaj-a06c4f3d0a1b4a269e97432bb1b6d0be |
|---|---|
| record_format |
Article |
| spelling |
doaj-a06c4f3d0a1b4a269e97432bb1b6d0be2021-02-11T06:18:20ZengFrontiers Media S.A.Frontiers in Neuroscience1662-453X2021-02-011510.3389/fnins.2021.628254628254Coupled VO2 Oscillators Circuit as Analog First Layer Filter in Convolutional Neural NetworksElisabetta Corti0Joaquin Antonio Cornejo Jimenez1Kham M. Niang2John Robertson3Kirsten E. Moselund4Bernd Gotsmann5Adrian M. Ionescu6Siegfried Karg7IBM Research Zürich, Rüschlikon, SwitzerlandIBM Research Zürich, Rüschlikon, SwitzerlandDepartment of Engineering, University of Cambridge, Cambridge, United KingdomDepartment of Engineering, University of Cambridge, Cambridge, United KingdomIBM Research Zürich, Rüschlikon, SwitzerlandIBM Research Zürich, Rüschlikon, SwitzerlandNanoelectronic Devices Laboratory, École Polytechnique Fédérale de Lausanne, Lausanne, SwitzerlandIBM Research Zürich, Rüschlikon, SwitzerlandIn this work we present an in-memory computing platform based on coupled VO2 oscillators fabricated in a crossbar configuration on silicon. Compared to existing platforms, the crossbar configuration promises significant improvements in terms of area density and oscillation frequency. Further, the crossbar devices exhibit low variability and extended reliability, hence, enabling experiments on 4-coupled oscillator. We demonstrate the neuromorphic computing capabilities using the phase relation of the oscillators. As an application, we propose to replace digital filtering operation in a convolutional neural network with oscillating circuits. The concept is tested with a VGG13 architecture on the MNIST dataset, achieving performances of 95% in the recognition task.https://www.frontiersin.org/articles/10.3389/fnins.2021.628254/fulloscillatory neural networkvanadium dioxidephase-encodingconvolutional neural networkspattern recognitionrelaxation oscillators |
| collection |
DOAJ |
| language |
English |
| format |
Article |
| sources |
DOAJ |
| author |
Elisabetta Corti Joaquin Antonio Cornejo Jimenez Kham M. Niang John Robertson Kirsten E. Moselund Bernd Gotsmann Adrian M. Ionescu Siegfried Karg |
| spellingShingle |
Elisabetta Corti Joaquin Antonio Cornejo Jimenez Kham M. Niang John Robertson Kirsten E. Moselund Bernd Gotsmann Adrian M. Ionescu Siegfried Karg Coupled VO2 Oscillators Circuit as Analog First Layer Filter in Convolutional Neural Networks Frontiers in Neuroscience oscillatory neural network vanadium dioxide phase-encoding convolutional neural networks pattern recognition relaxation oscillators |
| author_facet |
Elisabetta Corti Joaquin Antonio Cornejo Jimenez Kham M. Niang John Robertson Kirsten E. Moselund Bernd Gotsmann Adrian M. Ionescu Siegfried Karg |
| author_sort |
Elisabetta Corti |
| title |
Coupled VO2 Oscillators Circuit as Analog First Layer Filter in Convolutional Neural Networks |
| title_short |
Coupled VO2 Oscillators Circuit as Analog First Layer Filter in Convolutional Neural Networks |
| title_full |
Coupled VO2 Oscillators Circuit as Analog First Layer Filter in Convolutional Neural Networks |
| title_fullStr |
Coupled VO2 Oscillators Circuit as Analog First Layer Filter in Convolutional Neural Networks |
| title_full_unstemmed |
Coupled VO2 Oscillators Circuit as Analog First Layer Filter in Convolutional Neural Networks |
| title_sort |
coupled vo2 oscillators circuit as analog first layer filter in convolutional neural networks |
| publisher |
Frontiers Media S.A. |
| series |
Frontiers in Neuroscience |
| issn |
1662-453X |
| publishDate |
2021-02-01 |
| description |
In this work we present an in-memory computing platform based on coupled VO2 oscillators fabricated in a crossbar configuration on silicon. Compared to existing platforms, the crossbar configuration promises significant improvements in terms of area density and oscillation frequency. Further, the crossbar devices exhibit low variability and extended reliability, hence, enabling experiments on 4-coupled oscillator. We demonstrate the neuromorphic computing capabilities using the phase relation of the oscillators. As an application, we propose to replace digital filtering operation in a convolutional neural network with oscillating circuits. The concept is tested with a VGG13 architecture on the MNIST dataset, achieving performances of 95% in the recognition task. |
| topic |
oscillatory neural network vanadium dioxide phase-encoding convolutional neural networks pattern recognition relaxation oscillators |
| url |
https://www.frontiersin.org/articles/10.3389/fnins.2021.628254/full |
| work_keys_str_mv |
AT elisabettacorti coupledvo2oscillatorscircuitasanalogfirstlayerfilterinconvolutionalneuralnetworks AT joaquinantoniocornejojimenez coupledvo2oscillatorscircuitasanalogfirstlayerfilterinconvolutionalneuralnetworks AT khammniang coupledvo2oscillatorscircuitasanalogfirstlayerfilterinconvolutionalneuralnetworks AT johnrobertson coupledvo2oscillatorscircuitasanalogfirstlayerfilterinconvolutionalneuralnetworks AT kirstenemoselund coupledvo2oscillatorscircuitasanalogfirstlayerfilterinconvolutionalneuralnetworks AT berndgotsmann coupledvo2oscillatorscircuitasanalogfirstlayerfilterinconvolutionalneuralnetworks AT adrianmionescu coupledvo2oscillatorscircuitasanalogfirstlayerfilterinconvolutionalneuralnetworks AT siegfriedkarg coupledvo2oscillatorscircuitasanalogfirstlayerfilterinconvolutionalneuralnetworks |
| _version_ |
1724274504930164736 |