Perspective: Stochastic magnetic devices for cognitive computing

Kaushik Roy, Abhronil Sengupta, Yong Shim

Research output: Contribution to journalArticle

5 Citations (Scopus)

Abstract

Stochastic switching of nanomagnets can potentially enable probabilistic cognitive hardware consisting of noisy neural and synaptic components. Furthermore, computational paradigms inspired from the Ising computing model require stochasticity for achieving near-optimality in solutions to various types of combinatorial optimization problems such as the Graph Coloring Problem or the Travelling Salesman Problem. Achieving optimal solutions in such problems are computationally exhaustive and requires natural annealing to arrive at the near-optimal solutions. Stochastic switching of devices also finds use in applications involving Deep Belief Networks and Bayesian Inference. In this article, we provide a multi-disciplinary perspective across the stack of devices, circuits, and algorithms to illustrate how the stochastic switching dynamics of spintronic devices in the presence of thermal noise can provide a direct mapping to the computational units of such probabilistic intelligent systems.

Original languageEnglish (US)
Article number210901
JournalJournal of Applied Physics
Volume123
Issue number21
DOIs
StatePublished - Jun 7 2018

Fingerprint

belief networks
traveling salesman problem
thermal noise
inference
hardware
optimization
annealing

All Science Journal Classification (ASJC) codes

  • Physics and Astronomy(all)

Cite this

@article{8c711c1d78954329b47ffcf2c45d30ce,
title = "Perspective: Stochastic magnetic devices for cognitive computing",
abstract = "Stochastic switching of nanomagnets can potentially enable probabilistic cognitive hardware consisting of noisy neural and synaptic components. Furthermore, computational paradigms inspired from the Ising computing model require stochasticity for achieving near-optimality in solutions to various types of combinatorial optimization problems such as the Graph Coloring Problem or the Travelling Salesman Problem. Achieving optimal solutions in such problems are computationally exhaustive and requires natural annealing to arrive at the near-optimal solutions. Stochastic switching of devices also finds use in applications involving Deep Belief Networks and Bayesian Inference. In this article, we provide a multi-disciplinary perspective across the stack of devices, circuits, and algorithms to illustrate how the stochastic switching dynamics of spintronic devices in the presence of thermal noise can provide a direct mapping to the computational units of such probabilistic intelligent systems.",
author = "Kaushik Roy and Abhronil Sengupta and Yong Shim",
year = "2018",
month = "6",
day = "7",
doi = "10.1063/1.5020168",
language = "English (US)",
volume = "123",
journal = "Journal of Applied Physics",
issn = "0021-8979",
publisher = "American Institute of Physics Publising LLC",
number = "21",

}

Perspective : Stochastic magnetic devices for cognitive computing. / Roy, Kaushik; Sengupta, Abhronil; Shim, Yong.

In: Journal of Applied Physics, Vol. 123, No. 21, 210901, 07.06.2018.

Research output: Contribution to journalArticle

TY - JOUR

T1 - Perspective

T2 - Stochastic magnetic devices for cognitive computing

AU - Roy, Kaushik

AU - Sengupta, Abhronil

AU - Shim, Yong

PY - 2018/6/7

Y1 - 2018/6/7

N2 - Stochastic switching of nanomagnets can potentially enable probabilistic cognitive hardware consisting of noisy neural and synaptic components. Furthermore, computational paradigms inspired from the Ising computing model require stochasticity for achieving near-optimality in solutions to various types of combinatorial optimization problems such as the Graph Coloring Problem or the Travelling Salesman Problem. Achieving optimal solutions in such problems are computationally exhaustive and requires natural annealing to arrive at the near-optimal solutions. Stochastic switching of devices also finds use in applications involving Deep Belief Networks and Bayesian Inference. In this article, we provide a multi-disciplinary perspective across the stack of devices, circuits, and algorithms to illustrate how the stochastic switching dynamics of spintronic devices in the presence of thermal noise can provide a direct mapping to the computational units of such probabilistic intelligent systems.

AB - Stochastic switching of nanomagnets can potentially enable probabilistic cognitive hardware consisting of noisy neural and synaptic components. Furthermore, computational paradigms inspired from the Ising computing model require stochasticity for achieving near-optimality in solutions to various types of combinatorial optimization problems such as the Graph Coloring Problem or the Travelling Salesman Problem. Achieving optimal solutions in such problems are computationally exhaustive and requires natural annealing to arrive at the near-optimal solutions. Stochastic switching of devices also finds use in applications involving Deep Belief Networks and Bayesian Inference. In this article, we provide a multi-disciplinary perspective across the stack of devices, circuits, and algorithms to illustrate how the stochastic switching dynamics of spintronic devices in the presence of thermal noise can provide a direct mapping to the computational units of such probabilistic intelligent systems.

UR - http://www.scopus.com/inward/record.url?scp=85048201504&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85048201504&partnerID=8YFLogxK

U2 - 10.1063/1.5020168

DO - 10.1063/1.5020168

M3 - Article

AN - SCOPUS:85048201504

VL - 123

JO - Journal of Applied Physics

JF - Journal of Applied Physics

SN - 0021-8979

IS - 21

M1 - 210901

ER -