Performance analysis and benchmarking of all-spin spiking neural networks (Special session paper)

Abhronil Sengupta, Aayush Ankit, Kaushik Roy

Research output: Chapter in Book/Report/Conference proceedingConference contribution

6 Citations (Scopus)

Abstract

Spiking Neural Network based brain-inspired computing paradigms are becoming increasingly popular tools for various cognitive tasks. The sparse event-driven processing capability enabled by such networks can be potentially appealing for implementation of low-power neural computing platforms. However, the parallel and memory-intensive computations involved in such algorithms is in complete contrast to the sequential fetch, decode, execute cycles of conventional von-Neumann processors. Recent proposals have investigated the design of spintronic 'in-memory' crossbar based computing architectures driving 'spin neurons' that can potentially alleviate the memory-access bottleneck of CMOS based systems and simultaneously offer the prospect of low-power inner product computations. In this article, we perform a rigorous system-level simulation study of such All-Spin Spiking Neural Networks on a benchmark suite of 6 recognition problems ranging in network complexity from 10k-7.4M synapses and 195-9.2k neurons. System level simulations indicate that the proposed spintronic architecture can potentially achieve ∼1292× energy efficiency and ∼ 235× speedup on average over the benchmark suite in comparison to an optimized CMOS implementation at 45nm technology node.

Original languageEnglish (US)
Title of host publication2017 International Joint Conference on Neural Networks, IJCNN 2017 - Proceedings
PublisherInstitute of Electrical and Electronics Engineers Inc.
Pages4557-4563
Number of pages7
ISBN (Electronic)9781509061815
DOIs
StatePublished - Jun 30 2017
Event2017 International Joint Conference on Neural Networks, IJCNN 2017 - Anchorage, United States
Duration: May 14 2017May 19 2017

Publication series

NameProceedings of the International Joint Conference on Neural Networks
Volume2017-May

Conference

Conference2017 International Joint Conference on Neural Networks, IJCNN 2017
CountryUnited States
CityAnchorage
Period5/14/175/19/17

Fingerprint

Benchmarking
Magnetoelectronics
Neural networks
Data storage equipment
Neurons
Energy efficiency
Brain
Processing

All Science Journal Classification (ASJC) codes

  • Software
  • Artificial Intelligence

Cite this

Sengupta, A., Ankit, A., & Roy, K. (2017). Performance analysis and benchmarking of all-spin spiking neural networks (Special session paper). In 2017 International Joint Conference on Neural Networks, IJCNN 2017 - Proceedings (pp. 4557-4563). [7966434] (Proceedings of the International Joint Conference on Neural Networks; Vol. 2017-May). Institute of Electrical and Electronics Engineers Inc.. https://doi.org/10.1109/IJCNN.2017.7966434
Sengupta, Abhronil ; Ankit, Aayush ; Roy, Kaushik. / Performance analysis and benchmarking of all-spin spiking neural networks (Special session paper). 2017 International Joint Conference on Neural Networks, IJCNN 2017 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2017. pp. 4557-4563 (Proceedings of the International Joint Conference on Neural Networks).
@inproceedings{5c7b2011ea974777a2333e7fc0a3e067,
title = "Performance analysis and benchmarking of all-spin spiking neural networks (Special session paper)",
abstract = "Spiking Neural Network based brain-inspired computing paradigms are becoming increasingly popular tools for various cognitive tasks. The sparse event-driven processing capability enabled by such networks can be potentially appealing for implementation of low-power neural computing platforms. However, the parallel and memory-intensive computations involved in such algorithms is in complete contrast to the sequential fetch, decode, execute cycles of conventional von-Neumann processors. Recent proposals have investigated the design of spintronic 'in-memory' crossbar based computing architectures driving 'spin neurons' that can potentially alleviate the memory-access bottleneck of CMOS based systems and simultaneously offer the prospect of low-power inner product computations. In this article, we perform a rigorous system-level simulation study of such All-Spin Spiking Neural Networks on a benchmark suite of 6 recognition problems ranging in network complexity from 10k-7.4M synapses and 195-9.2k neurons. System level simulations indicate that the proposed spintronic architecture can potentially achieve ∼1292× energy efficiency and ∼ 235× speedup on average over the benchmark suite in comparison to an optimized CMOS implementation at 45nm technology node.",
author = "Abhronil Sengupta and Aayush Ankit and Kaushik Roy",
year = "2017",
month = "6",
day = "30",
doi = "10.1109/IJCNN.2017.7966434",
language = "English (US)",
series = "Proceedings of the International Joint Conference on Neural Networks",
publisher = "Institute of Electrical and Electronics Engineers Inc.",
pages = "4557--4563",
booktitle = "2017 International Joint Conference on Neural Networks, IJCNN 2017 - Proceedings",
address = "United States",

}

Sengupta, A, Ankit, A & Roy, K 2017, Performance analysis and benchmarking of all-spin spiking neural networks (Special session paper). in 2017 International Joint Conference on Neural Networks, IJCNN 2017 - Proceedings., 7966434, Proceedings of the International Joint Conference on Neural Networks, vol. 2017-May, Institute of Electrical and Electronics Engineers Inc., pp. 4557-4563, 2017 International Joint Conference on Neural Networks, IJCNN 2017, Anchorage, United States, 5/14/17. https://doi.org/10.1109/IJCNN.2017.7966434

Performance analysis and benchmarking of all-spin spiking neural networks (Special session paper). / Sengupta, Abhronil; Ankit, Aayush; Roy, Kaushik.

2017 International Joint Conference on Neural Networks, IJCNN 2017 - Proceedings. Institute of Electrical and Electronics Engineers Inc., 2017. p. 4557-4563 7966434 (Proceedings of the International Joint Conference on Neural Networks; Vol. 2017-May).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

TY - GEN

T1 - Performance analysis and benchmarking of all-spin spiking neural networks (Special session paper)

AU - Sengupta, Abhronil

AU - Ankit, Aayush

AU - Roy, Kaushik

PY - 2017/6/30

Y1 - 2017/6/30

N2 - Spiking Neural Network based brain-inspired computing paradigms are becoming increasingly popular tools for various cognitive tasks. The sparse event-driven processing capability enabled by such networks can be potentially appealing for implementation of low-power neural computing platforms. However, the parallel and memory-intensive computations involved in such algorithms is in complete contrast to the sequential fetch, decode, execute cycles of conventional von-Neumann processors. Recent proposals have investigated the design of spintronic 'in-memory' crossbar based computing architectures driving 'spin neurons' that can potentially alleviate the memory-access bottleneck of CMOS based systems and simultaneously offer the prospect of low-power inner product computations. In this article, we perform a rigorous system-level simulation study of such All-Spin Spiking Neural Networks on a benchmark suite of 6 recognition problems ranging in network complexity from 10k-7.4M synapses and 195-9.2k neurons. System level simulations indicate that the proposed spintronic architecture can potentially achieve ∼1292× energy efficiency and ∼ 235× speedup on average over the benchmark suite in comparison to an optimized CMOS implementation at 45nm technology node.

AB - Spiking Neural Network based brain-inspired computing paradigms are becoming increasingly popular tools for various cognitive tasks. The sparse event-driven processing capability enabled by such networks can be potentially appealing for implementation of low-power neural computing platforms. However, the parallel and memory-intensive computations involved in such algorithms is in complete contrast to the sequential fetch, decode, execute cycles of conventional von-Neumann processors. Recent proposals have investigated the design of spintronic 'in-memory' crossbar based computing architectures driving 'spin neurons' that can potentially alleviate the memory-access bottleneck of CMOS based systems and simultaneously offer the prospect of low-power inner product computations. In this article, we perform a rigorous system-level simulation study of such All-Spin Spiking Neural Networks on a benchmark suite of 6 recognition problems ranging in network complexity from 10k-7.4M synapses and 195-9.2k neurons. System level simulations indicate that the proposed spintronic architecture can potentially achieve ∼1292× energy efficiency and ∼ 235× speedup on average over the benchmark suite in comparison to an optimized CMOS implementation at 45nm technology node.

UR - http://www.scopus.com/inward/record.url?scp=85031034828&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85031034828&partnerID=8YFLogxK

U2 - 10.1109/IJCNN.2017.7966434

DO - 10.1109/IJCNN.2017.7966434

M3 - Conference contribution

AN - SCOPUS:85031034828

T3 - Proceedings of the International Joint Conference on Neural Networks

SP - 4557

EP - 4563

BT - 2017 International Joint Conference on Neural Networks, IJCNN 2017 - Proceedings

PB - Institute of Electrical and Electronics Engineers Inc.

ER -

Sengupta A, Ankit A, Roy K. Performance analysis and benchmarking of all-spin spiking neural networks (Special session paper). In 2017 International Joint Conference on Neural Networks, IJCNN 2017 - Proceedings. Institute of Electrical and Electronics Engineers Inc. 2017. p. 4557-4563. 7966434. (Proceedings of the International Joint Conference on Neural Networks). https://doi.org/10.1109/IJCNN.2017.7966434