Higher-Accuracy Photonic Neural Networks via Duplication Schemes for Noise Reduction

Gianluca Kosmella; Jaron Sanders; Bin Shi; Ripalta Stabile

doi:10.1109/PSC57974.2023.10297256

Higher-Accuracy Photonic Neural Networks via Duplication Schemes for Noise Reduction

Gianluca Kosmella, Jaron Sanders, Bin Shi, Ripalta Stabile

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

1 Downloads (Pure)

Abstract

We present how feasible duplication schemes for reducing noise in optical neural networks achieve accuracy gains when compared to implementations without duplication. Performance gains are 7.4% at a practical chip size, and noise
can be negated completely in a many-duplication regime.

Original language	English
Title of host publication	2023 International Conference on Photonics in Switching and Computing (PSC)
Publisher	Institute of Electrical and Electronics Engineers
Pages	1-4
Number of pages	4
ISBN (Electronic)	979-8-3503-2370-2
DOIs	https://doi.org/10.1109/PSC57974.2023.10297256
Publication status	Published - 2 Nov 2023
Event	2023 International Conference on Photonics in Switching and Computing, PSC 2023 - Mantova, Italy Duration: 26 Sept 2023 → 29 Sept 2023

Conference

Conference	2023 International Conference on Photonics in Switching and Computing, PSC 2023
Abbreviated title	PSC 2023
Country/Territory	Italy
City	Mantova
Period	26/09/23 → 29/09/23

Funding

This research was supported by the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement no. 945045, and by the NWO Gravitation project NETWORKS under grant no. 024.002.003.

Funders	Funder number
European Union's Horizon 2020 - Research and Innovation Framework Programme
Marie Skłodowska‐Curie	945045
Nederlandse Organisatie voor Wetenschappelijk Onderzoek	024.002.003

Keywords

Law of Large Numbers
Optical Neural Networks
Universal Approximation

Access to Document

10.1109/PSC57974.2023.10297256

Cite this

@inproceedings{0a2a2808f0914eb998cb5659f73f8fb3,

title = "Higher-Accuracy Photonic Neural Networks via Duplication Schemes for Noise Reduction",

abstract = "We present how feasible duplication schemes for reducing noise in optical neural networks achieve accuracy gains when compared to implementations without duplication. Performance gains are 7.4% at a practical chip size, and noisecan be negated completely in a many-duplication regime.",

keywords = "Law of Large Numbers, Optical Neural Networks, Universal Approximation",

author = "Gianluca Kosmella and Jaron Sanders and Bin Shi and Ripalta Stabile",

year = "2023",

month = nov,

day = "2",

doi = "10.1109/PSC57974.2023.10297256",

language = "English",

pages = "1--4",

booktitle = "2023 International Conference on Photonics in Switching and Computing (PSC)",

publisher = "Institute of Electrical and Electronics Engineers",

address = "United States",

note = "2023 International Conference on Photonics in Switching and Computing, PSC 2023, PSC 2023 ; Conference date: 26-09-2023 Through 29-09-2023",

}

Kosmella, G , Sanders, J , Shi, B & Stabile, R 2023, Higher-Accuracy Photonic Neural Networks via Duplication Schemes for Noise Reduction. in 2023 International Conference on Photonics in Switching and Computing (PSC)., 10297256, Institute of Electrical and Electronics Engineers, pp. 1-4, 2023 International Conference on Photonics in Switching and Computing, PSC 2023, Mantova, Italy, 26/09/23. https://doi.org/10.1109/PSC57974.2023.10297256

Higher-Accuracy Photonic Neural Networks via Duplication Schemes for Noise Reduction. / Kosmella, Gianluca ; Sanders, Jaron ; Shi, Bin et al.
2023 International Conference on Photonics in Switching and Computing (PSC). Institute of Electrical and Electronics Engineers, 2023. p. 1-4 10297256.

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Academic › peer-review

TY - GEN

T1 - Higher-Accuracy Photonic Neural Networks via Duplication Schemes for Noise Reduction

AU - Kosmella, Gianluca

AU - Sanders, Jaron

AU - Shi, Bin

AU - Stabile, Ripalta

PY - 2023/11/2

Y1 - 2023/11/2

N2 - We present how feasible duplication schemes for reducing noise in optical neural networks achieve accuracy gains when compared to implementations without duplication. Performance gains are 7.4% at a practical chip size, and noisecan be negated completely in a many-duplication regime.

AB - We present how feasible duplication schemes for reducing noise in optical neural networks achieve accuracy gains when compared to implementations without duplication. Performance gains are 7.4% at a practical chip size, and noisecan be negated completely in a many-duplication regime.

KW - Law of Large Numbers

KW - Optical Neural Networks

KW - Universal Approximation

UR - http://www.scopus.com/inward/record.url?scp=85178510540&partnerID=8YFLogxK

U2 - 10.1109/PSC57974.2023.10297256

DO - 10.1109/PSC57974.2023.10297256

M3 - Conference contribution

SP - 1

EP - 4

BT - 2023 International Conference on Photonics in Switching and Computing (PSC)

PB - Institute of Electrical and Electronics Engineers

T2 - 2023 International Conference on Photonics in Switching and Computing, PSC 2023

Y2 - 26 September 2023 through 29 September 2023

ER -

Higher-Accuracy Photonic Neural Networks via Duplication Schemes for Noise Reduction

Abstract

Conference

Funding

Keywords

Access to Document

Other files and links

Fingerprint

Cite this