Knowledge Distillation Applied to Optical Channel Equalization:: Solving the Parallelization Problem of Recurrent Connection

Sasipim Srivallapanondh; Pedro J. Freire; Bernhard Spinnler; Nelson Costa; Antonio Napoli; Sergei K. Turitsyn; Jaroslaw E. Prilepsky

doi:10.23919/ofc49934.2023.10116725

Knowledge Distillation Applied to Optical Channel Equalization: Solving the Parallelization Problem of Recurrent Connection

Sasipim Srivallapanondh, Pedro J. Freire, Bernhard Spinnler, Nelson Costa, Antonio Napoli, Sergei K. Turitsyn, Jaroslaw E. Prilepsky

Research output: Chapter in Book/Published conference output › Conference publication

Abstract

To circumvent the non-parallelizability of recurrent neural network-based equalizers, we propose knowledge distillation to recast the RNN into a parallelizable feedforward structure. The latter shows 38% latency decrease, while impacting the Q-factor by only 0.5 dB.

Original language	English
Title of host publication	Proceedings of 2023 Optical Fiber Communications Conference and Exhibition (OFC)
Publisher	IEEE
ISBN (Electronic)	9781957171180
DOIs	https://doi.org/10.23919/ofc49934.2023.10116725
Publication status	Published - 5 Mar 2023
Event	2023 Optical Fiber Communications Conference and Exhibition (OFC) - Duration: 5 Mar 2023 → 9 Mar 2023

Publication series

Name	2023 Optical Fiber Communications Conference and Exhibition (OFC)
Publisher	IEEE

Conference

Conference	2023 Optical Fiber Communications Conference and Exhibition (OFC)
Abbreviated title	OFC
Period	5/03/23 → 9/03/23

Bibliographical note

Funding Information:
Acknowledgements: This work has received funding from the EU Horizon 2020 program under the Marie Skłodowska-Curie grant agreement No. 956713 (MENTOR) and 813144 (REAL-NET). SKT acknowledges the support of the EPSRC project TRANSNET (EP/R035342/1).

Publisher Copyright:
© 2023 The Author(s).

Access to Document

10.23919/ofc49934.2023.10116725

Cite this

Srivallapanondh, S., Freire, P. J., Spinnler, B., Costa, N., Napoli, A., Turitsyn, S. K., & Prilepsky, J. E. (2023). Knowledge Distillation Applied to Optical Channel Equalization: Solving the Parallelization Problem of Recurrent Connection. In Proceedings of 2023 Optical Fiber Communications Conference and Exhibition (OFC) (2023 Optical Fiber Communications Conference and Exhibition (OFC)). IEEE. https://doi.org/10.23919/ofc49934.2023.10116725

@inproceedings{af7349a9cfe24538a0dc8fecc1754bbd,

title = "Knowledge Distillation Applied to Optical Channel Equalization:: Solving the Parallelization Problem of Recurrent Connection",

abstract = "To circumvent the non-parallelizability of recurrent neural network-based equalizers, we propose knowledge distillation to recast the RNN into a parallelizable feedforward structure. The latter shows 38% latency decrease, while impacting the Q-factor by only 0.5 dB.",

author = "Sasipim Srivallapanondh and Freire, {Pedro J.} and Bernhard Spinnler and Nelson Costa and Antonio Napoli and Turitsyn, {Sergei K.} and Prilepsky, {Jaroslaw E.}",

note = "Funding Information: Acknowledgements: This work has received funding from the EU Horizon 2020 program under the Marie Sk{\l}odowska-Curie grant agreement No. 956713 (MENTOR) and 813144 (REAL-NET). SKT acknowledges the support of the EPSRC project TRANSNET (EP/R035342/1). Publisher Copyright: {\textcopyright} 2023 The Author(s).; 2023 Optical Fiber Communications Conference and Exhibition (OFC), OFC ; Conference date: 05-03-2023 Through 09-03-2023",

year = "2023",

month = mar,

day = "5",

doi = "10.23919/ofc49934.2023.10116725",

language = "English",

series = "2023 Optical Fiber Communications Conference and Exhibition (OFC)",

publisher = "IEEE",

booktitle = "Proceedings of 2023 Optical Fiber Communications Conference and Exhibition (OFC)",

address = "United States",

}

Srivallapanondh, S, Freire, PJ, Spinnler, B, Costa, N, Napoli, A, Turitsyn, SK & Prilepsky, JE 2023, Knowledge Distillation Applied to Optical Channel Equalization: Solving the Parallelization Problem of Recurrent Connection. in Proceedings of 2023 Optical Fiber Communications Conference and Exhibition (OFC). 2023 Optical Fiber Communications Conference and Exhibition (OFC), IEEE, 2023 Optical Fiber Communications Conference and Exhibition (OFC), 5/03/23. https://doi.org/10.23919/ofc49934.2023.10116725

Knowledge Distillation Applied to Optical Channel Equalization: Solving the Parallelization Problem of Recurrent Connection. / Srivallapanondh, Sasipim; Freire, Pedro J.; Spinnler, Bernhard et al.
Proceedings of 2023 Optical Fiber Communications Conference and Exhibition (OFC). IEEE, 2023. (2023 Optical Fiber Communications Conference and Exhibition (OFC)).

Research output: Chapter in Book/Published conference output › Conference publication

TY - GEN

T1 - Knowledge Distillation Applied to Optical Channel Equalization:

T2 - 2023 Optical Fiber Communications Conference and Exhibition (OFC)

AU - Srivallapanondh, Sasipim

AU - Freire, Pedro J.

AU - Spinnler, Bernhard

AU - Costa, Nelson

AU - Napoli, Antonio

AU - Turitsyn, Sergei K.

AU - Prilepsky, Jaroslaw E.

N1 - Funding Information: Acknowledgements: This work has received funding from the EU Horizon 2020 program under the Marie Skłodowska-Curie grant agreement No. 956713 (MENTOR) and 813144 (REAL-NET). SKT acknowledges the support of the EPSRC project TRANSNET (EP/R035342/1). Publisher Copyright: © 2023 The Author(s).

PY - 2023/3/5

Y1 - 2023/3/5

N2 - To circumvent the non-parallelizability of recurrent neural network-based equalizers, we propose knowledge distillation to recast the RNN into a parallelizable feedforward structure. The latter shows 38% latency decrease, while impacting the Q-factor by only 0.5 dB.

AB - To circumvent the non-parallelizability of recurrent neural network-based equalizers, we propose knowledge distillation to recast the RNN into a parallelizable feedforward structure. The latter shows 38% latency decrease, while impacting the Q-factor by only 0.5 dB.

UR - https://ieeexplore.ieee.org/document/10116725

UR - http://www.scopus.com/inward/record.url?scp=85161348297&partnerID=8YFLogxK

U2 - 10.23919/ofc49934.2023.10116725

DO - 10.23919/ofc49934.2023.10116725

M3 - Conference publication

T3 - 2023 Optical Fiber Communications Conference and Exhibition (OFC)

BT - Proceedings of 2023 Optical Fiber Communications Conference and Exhibition (OFC)

PB - IEEE

Y2 - 5 March 2023 through 9 March 2023

ER -

Srivallapanondh S, Freire PJ, Spinnler B, Costa N, Napoli A, Turitsyn SK et al. Knowledge Distillation Applied to Optical Channel Equalization: Solving the Parallelization Problem of Recurrent Connection. In Proceedings of 2023 Optical Fiber Communications Conference and Exhibition (OFC). IEEE. 2023. (2023 Optical Fiber Communications Conference and Exhibition (OFC)). doi: 10.23919/ofc49934.2023.10116725