Trainable dynamical masking for readout-free optical computing

S. Bogdanov; E. Manuylovich; S. K. Turitsyn

doi:10.1364/ol.566999

Trainable dynamical masking for readout-free optical computing

S. Bogdanov, E. Manuylovich, S. K. Turitsyn

Aston University

Research output: Contribution to journal › Article › peer-review

4 Downloads (Pure)

Abstract

Nonlinear systems, transforming an input signal into a high-dimensional output feature space, can be used for non-conventional computing. This approach, however, requires a change of system parameters during training rather than coefficients in a software program. We propose here to use available off-the-shelf high-speed optical communication devices and technologies to implement a trainable dynamical mask in addition to or even instead of the traditional readout layer for extreme learning machine-based computing. The computational potential of the proposed approach is demonstrated in numerical simulations with both regression and time series prediction tasks.

Original language	English
Article number	5554
Pages (from-to)	5554-5557
Number of pages	4
Journal	Optics Letters
Volume	50
Issue number	18
Early online date	2 Sept 2025
DOIs	https://doi.org/10.1364/ol.566999
Publication status	Published - 15 Sept 2025

Bibliographical note

Published by Optica Publishing Group under the terms of the Creative Commons Attribution 4.0 License. Further distribution of this work must maintain attribution to the author(s) and the published article’s title, journal citation, and DOI.

Funding

Funding acknowledged: Engineering and Physical Sciences Research Council (EP/W002868/1).

Funders	Funder number
Engineering and Physical Sciences Research Council	EP/W002868/1
Engineering and Physical Sciences Research Council

Access to Document

10.1364/ol.566999Licence: CC BY 4.0

Bogdanov Manuylovich and Turitsyn_ Trainable dynamical masking for readout-free optical computing
Published by Optica Publishing Group under the terms of the Creative Commons Attribution 4.0 License. Further distribution of this work must maintain attribution to the author(s) and the published article’s title, journal citation, and DOI.
Final published version, 2.02 MBLicence: CC BY 4.0

Cite this

@article{e60aafd066e042c98cb57f8382509f05,

title = "Trainable dynamical masking for readout-free optical computing",

abstract = "Nonlinear systems, transforming an input signal into a high-dimensional output feature space, can be used for non-conventional computing. This approach, however, requires a change of system parameters during training rather than coefficients in a software program. We propose here to use available off-the-shelf high-speed optical communication devices and technologies to implement a trainable dynamical mask in addition to or even instead of the traditional readout layer for extreme learning machine-based computing. The computational potential of the proposed approach is demonstrated in numerical simulations with both regression and time series prediction tasks.",

author = "S. Bogdanov and E. Manuylovich and Turitsyn, {S. K.}",

note = "Published by Optica Publishing Group under the terms of the Creative Commons Attribution 4.0 License. Further distribution of this work must maintain attribution to the author(s) and the published article{\textquoteright}s title, journal citation, and DOI.",

year = "2025",

month = sep,

day = "15",

doi = "10.1364/ol.566999",

language = "English",

volume = "50",

pages = "5554--5557",

journal = "Optics Letters",

issn = "0146-9592",

publisher = "Optica Publishing Group (formerly OSA)",

number = "18",

}

TY - JOUR

T1 - Trainable dynamical masking for readout-free optical computing

AU - Bogdanov, S.

AU - Manuylovich, E.

AU - Turitsyn, S. K.

N1 - Published by Optica Publishing Group under the terms of the Creative Commons Attribution 4.0 License. Further distribution of this work must maintain attribution to the author(s) and the published article’s title, journal citation, and DOI.

PY - 2025/9/15

Y1 - 2025/9/15

N2 - Nonlinear systems, transforming an input signal into a high-dimensional output feature space, can be used for non-conventional computing. This approach, however, requires a change of system parameters during training rather than coefficients in a software program. We propose here to use available off-the-shelf high-speed optical communication devices and technologies to implement a trainable dynamical mask in addition to or even instead of the traditional readout layer for extreme learning machine-based computing. The computational potential of the proposed approach is demonstrated in numerical simulations with both regression and time series prediction tasks.

AB - Nonlinear systems, transforming an input signal into a high-dimensional output feature space, can be used for non-conventional computing. This approach, however, requires a change of system parameters during training rather than coefficients in a software program. We propose here to use available off-the-shelf high-speed optical communication devices and technologies to implement a trainable dynamical mask in addition to or even instead of the traditional readout layer for extreme learning machine-based computing. The computational potential of the proposed approach is demonstrated in numerical simulations with both regression and time series prediction tasks.

UR - https://opg.optica.org/ol/fulltext.cfm?uri=ol-50-18-5554&id=576219

UR - http://www.scopus.com/inward/record.url?scp=105015079306&partnerID=8YFLogxK

U2 - 10.1364/ol.566999

DO - 10.1364/ol.566999

M3 - Article

SN - 0146-9592

VL - 50

SP - 5554

EP - 5557

JO - Optics Letters

JF - Optics Letters

IS - 18

M1 - 5554

ER -

Trainable dynamical masking for readout-free optical computing

Abstract

Bibliographical note

Funding

Access to Document

Other files and links

Fingerprint

Cite this