Application of adaptive activation unit based on injection-locked lasers in machine learning tasks

Jasna Crnjanski; Mladen Banovic; Isidora Teofilovic; Marko Krstic; Dejan Gvozdic

doi:10.1109/SUM57928.2023.10224462

Application of adaptive activation unit based on injection-locked lasers in machine learning tasks

Jasna Crnjanski, Mladen Banovic, Isidora Teofilovic, Marko Krstic, Dejan Gvozdic

University of Belgrade

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

Abstract

The expressiveness and learning capabilities of neural networks depend on the activation functions nonlinearity. We present a concept of photonic device with nonlinear transfer function which can be flexibly reconfigured to optimize the performance of different machine learning tasks, such as classification or time-series prediction.

Original language	English
Title of host publication	Proceedings of 2023 IEEE Photonics Society Summer Topicals Meeting Series
Number of pages	2
Publisher	IEEE
Publication date	2023
Pages	1-2
ISBN (Print)	979-8-3503-4721-0
DOIs	https://doi.org/10.1109/SUM57928.2023.10224462
Publication status	Published - 2023
Event	2023 IEEE Photonics Society Summer Topicals Meeting Series (SUM) - RG Naxos Hotel, Giardini-Naxos, Italy Duration: 17 Jul 2023 → 19 Jul 2023

Conference

Conference	2023 IEEE Photonics Society Summer Topicals Meeting Series (SUM)
Location	RG Naxos Hotel
Country/Territory	Italy
City	Giardini-Naxos
Period	17/07/2023 → 19/07/2023

Access to Document

10.1109/SUM57928.2023.10224462

OpenUrl availability

Full text

Cite this

@inproceedings{f85bb98302b54b8c96f494eb8d5eba0e,

title = "Application of adaptive activation unit based on injection-locked lasers in machine learning tasks",

abstract = "The expressiveness and learning capabilities of neural networks depend on the activation functions nonlinearity. We present a concept of photonic device with nonlinear transfer function which can be flexibly reconfigured to optimize the performance of different machine learning tasks, such as classification or time-series prediction.",

author = "Jasna Crnjanski and Mladen Banovic and Isidora Teofilovic and Marko Krstic and Dejan Gvozdic",

year = "2023",

doi = "10.1109/SUM57928.2023.10224462",

language = "English",

isbn = "979-8-3503-4721-0",

pages = "1--2",

booktitle = "Proceedings of 2023 IEEE Photonics Society Summer Topicals Meeting Series",

publisher = "IEEE",

address = "United States",

note = "2023 IEEE Photonics Society Summer Topicals Meeting Series (SUM), SUM 2023 ; Conference date: 17-07-2023 Through 19-07-2023",

}

Crnjanski, J, Banovic, M, Teofilovic, I, Krstic, M & Gvozdic, D 2023, Application of adaptive activation unit based on injection-locked lasers in machine learning tasks. in Proceedings of 2023 IEEE Photonics Society Summer Topicals Meeting Series . IEEE, pp. 1-2, 2023 IEEE Photonics Society Summer Topicals Meeting Series (SUM), Giardini-Naxos, Italy, 17/07/2023. https://doi.org/10.1109/SUM57928.2023.10224462

Application of adaptive activation unit based on injection-locked lasers in machine learning tasks. / Crnjanski, Jasna; Banovic, Mladen; Teofilovic, Isidora et al.
Proceedings of 2023 IEEE Photonics Society Summer Topicals Meeting Series . IEEE, 2023. p. 1-2.

Research output: Chapter in Book/Report/Conference proceeding › Article in proceedings › Research › peer-review

TY - GEN

T1 - Application of adaptive activation unit based on injection-locked lasers in machine learning tasks

AU - Crnjanski, Jasna

AU - Banovic, Mladen

AU - Teofilovic, Isidora

AU - Krstic, Marko

AU - Gvozdic, Dejan

PY - 2023

Y1 - 2023

N2 - The expressiveness and learning capabilities of neural networks depend on the activation functions nonlinearity. We present a concept of photonic device with nonlinear transfer function which can be flexibly reconfigured to optimize the performance of different machine learning tasks, such as classification or time-series prediction.

AB - The expressiveness and learning capabilities of neural networks depend on the activation functions nonlinearity. We present a concept of photonic device with nonlinear transfer function which can be flexibly reconfigured to optimize the performance of different machine learning tasks, such as classification or time-series prediction.

U2 - 10.1109/SUM57928.2023.10224462

DO - 10.1109/SUM57928.2023.10224462

M3 - Article in proceedings

SN - 979-8-3503-4721-0

SP - 1

EP - 2

BT - Proceedings of 2023 IEEE Photonics Society Summer Topicals Meeting Series

PB - IEEE

T2 - 2023 IEEE Photonics Society Summer Topicals Meeting Series (SUM)

Y2 - 17 July 2023 through 19 July 2023

ER -

Application of adaptive activation unit based on injection-locked lasers in machine learning tasks

Abstract

Conference

Access to Document

OpenUrl availability

Fingerprint

Cite this