Multi-Scale Modeling and Simulation Flow for Oscillatory Neural Networks for Edge Computing

被引：6

作者：

Carapezzi, Stefania ^{[1
]}

Delacour, Corentin ^{[1
]}

Boschetto, Gabriele ^{[1
]}

Corti, Elisabetta ^{[2
]}

Abernot, Madeleine ^{[1
]}

Nejim, Ahmed ^{[3
]}

Gil, Thierry ^{[1
]}

Karg, Siegfried ^{[2
]}

Todri-Sanial, Aida ^{[1
]}

机构：

[1] Univ Montpellier, LIRMM, CNRS, Microelect Dept, Montpellier, France

[2] IBM Res Zurich, Dept Sci & Technol, Ruschlikon, Switzerland

[3] Silvaco Europe, Cambridge, England

来源：

2021 19TH IEEE INTERNATIONAL NEW CIRCUITS AND SYSTEMS CONFERENCE (NEWCAS) | 2021年

关键词：

Oscillatory neural networks (ONN); density functional theory (DFT); technology computer aided design (TCAD); circuit simulation; Internet-of-Things (IoT); edge artificial intelligence (edge AI); neuromorphic computing;

D O I：

10.1109/NEWCAS50681.2021.9462761

中图分类号：

TP3 [计算技术、计算机技术];

学科分类号：

0812 ;

摘要：

An oscillatory neural network (ONN) is a neuromorphic computing paradigm based on encoding of information into the phases of oscillators. In this paper we present an ONN whose elemental unit, the "neuron", is implemented through a beyondCMOS device based on vanadium dioxide (VO2). Such ONN technology provides ultra-low power solutions for performing pattern recognition tasks, and it is ideally suited for edge computing applications. However, exploring the groundwork of the beyond-CMOS ONN paradigm is mandatory premise for its industry-level exploitation. Such foundation entails the building of a holistic simulation flow from materials and devices to circuits, to allow assessment of ONN performance. In this work we report results of this advanced designing approach with special focus over the VO2 oscillator. This establishes the ground to scale up to evaluate beyond-CMOS ONN functionalities for pattern recognition.

引用

页数：5

共 16 条

[1] Tuning the work function of VO2(100) surface by Ag adsorption and incorporation: Insights from first-principles calculations [J].

Chen, Lanli ;

Wang, Xiaofang ;

Shi, Siqi ;

Cui, Yuanyuan ;

Luo, Hongjie ;

Gao, Yanfeng .

APPLIED SURFACE SCIENCE, 2016, 367 :507-517

[2] Calculation Evidence of Staged Mott and Peierls Transitions in VO2 Revealed by Mapping Reduced-Dimension Potential Energy Surface [J].

Chen, Shi ;

Liu, Jianjun ;

Luo, Hongjie ;

Gao, Yanfeng .

JOURNAL OF PHYSICAL CHEMISTRY LETTERS, 2015, 6 (18) :3650-3656

[3]

Corti E, 2018, 2018 IEEE INTERNATIONAL CONFERENCE ON REBOOTING COMPUTING (ICRC), P195, DOI 10.1109/ICRC.2018.8638626

[4] Modeling of temperature-dependent resistance in micro- and nanopolycrystalline VO2 thin films with random resistor networks [J].

Dai, Jun ;

Wang, Xingzhi ;

Huang, Ying ;

Yi, Xinjian .

OPTICAL ENGINEERING, 2008, 47 (03)

[5] Energy-Efficient Ferroelectric Field-Effect Transistor-Based Oscillators for Neuromorphic System Design [J].

Eslahi, Hossein ;

Hamilton, Tara J. ;

Khandelwal, Sourabh .

IEEE JOURNAL ON EXPLORATORY SOLID-STATE COMPUTATIONAL DEVICES AND CIRCUITS, 2020, 6 (02) :122-129

[6] Electronic properties of the VO2(011) surface:: density functional cluster calculations [J].

Haras, A ;

Witko, M ;

Salahub, DR ;

Hermann, K ;

Tokarz, R .

SURFACE SCIENCE, 2001, 491 (1-2) :77-87

[7] NEURAL NETWORKS AND PHYSICAL SYSTEMS WITH EMERGENT COLLECTIVE COMPUTATIONAL ABILITIES [J].

HOPFIELD, JJ .

PROCEEDINGS OF THE NATIONAL ACADEMY OF SCIENCES OF THE UNITED STATES OF AMERICA-BIOLOGICAL SCIENCES, 1982, 79 (08) :2554-2558

[8] Oscillatory neurocomputers with dynamic connectivity [J].

Hoppensteadt, FC ;

Izhikevich, EM .

PHYSICAL REVIEW LETTERS, 1999, 82 (14) :2983-2986

[9]

Kandel ER, PRINCIPLES NEURAL SC, V5th

[10] Joule Heating-Induced Metal-Insulator Transition in Epitaxial VO2/TiO2 Devices [J].

Li, Dasheng ;

Sharma, Abhishek A. ;

Gala, Darshil K. ;

Shukla, Nikhil ;

Paik, Hanjong ;

Datta, Suman ;

Schlom, Darrell G. ;

Bain, James A. ;

Skowronski, Marek .

ACS APPLIED MATERIALS & INTERFACES, 2016, 8 (20) :12908-12914

← 1 2 →