Multi-Step Span Loss Prediction in Optical Networks Using Multi-Head Attention Transformers

Span Loss is a pivotal characteristic of optical networks, and its accurate prediction enables adjustment for optimal performance and proactive monitoring. Deep learning models such as transformers, with their self-attention mechanism, have shown potential for various prediction tasks. In this study...

Full description

Saved in:
Bibliographic Details
Main Authors: Maryam Hedayatnejad, Yinqing Pei, David Boertjes, Dacian Demeter, Christian Desrosiers, Christine Tremblay
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Photonics Journal
Subjects:
Online Access:https://ieeexplore.ieee.org/document/11006412/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850123350843588608
author Maryam Hedayatnejad
Yinqing Pei
David Boertjes
Dacian Demeter
Christian Desrosiers
Christine Tremblay
author_facet Maryam Hedayatnejad
Yinqing Pei
David Boertjes
Dacian Demeter
Christian Desrosiers
Christine Tremblay
author_sort Maryam Hedayatnejad
collection DOAJ
description Span Loss is a pivotal characteristic of optical networks, and its accurate prediction enables adjustment for optimal performance and proactive monitoring. Deep learning models such as transformers, with their self-attention mechanism, have shown potential for various prediction tasks. In this study, we propose the Transformer-XL (Extra Long) model for single-step and multi-step forecasting, trained with field data. We report on models predicting span loss from 15 minutes to 5 days, using window sizes of 15 minutes to 10 days. The single-step model's average Absolute Maximum Error (AME) is better than the naive model by 2.13 dB and outperforms linear regression by 0.05–0.32 dB across different window sizes. Our single-step model also achieves better performance than the Recurrent Neural Network (RNN) with an AME improvement of 0.02 dB. The average AME of our multi-step model exceeds the naive model's performance by a range of 2.95-3.05 dB, linear regression by a substantial 0.02-0.15 dB and RNN by a range of 0.04-0.54 dB across different window sizes and forecast horizons. Based on Root Mean Square Error (RMSE), the single-step model performs better than the naive approach across various window sizes by 0.07 dB, achieves up to 0.07 dB improvement over linear regression, and delivers comparable results to RNN. Moreover, our multi-step model improves upon the naive approach with RMSE by 0.04 dB and RNN by 0.02 across various window sizes and forecast horizons. It also demonstrates a slight improvement over linear regression.
format Article
id doaj-art-3c868e9fd49f4f2c97ad7a820c185fe1
institution OA Journals
issn 1943-0655
language English
publishDate 2025-01-01
publisher IEEE
record_format Article
series IEEE Photonics Journal
spelling doaj-art-3c868e9fd49f4f2c97ad7a820c185fe12025-08-20T02:34:36ZengIEEEIEEE Photonics Journal1943-06552025-01-011731810.1109/JPHOT.2025.357122011006412Multi-Step Span Loss Prediction in Optical Networks Using Multi-Head Attention TransformersMaryam Hedayatnejad0https://orcid.org/0009-0003-4514-4737Yinqing Pei1David Boertjes2Dacian Demeter3Christian Desrosiers4https://orcid.org/0000-0002-9162-9650Christine Tremblay5https://orcid.org/0000-0002-9558-4824Network Technology Lab, Department of Electrical Engineering, École de technologie supérieure, Montréal, QC, CanadaCiena, Ottawa, ON, CanadaCiena, Ottawa, ON, CanadaTELUS Corp., Edmonton, BC, CanadaLaboratory for Imagery Vision and Artificial Intelligence, Department of Software, and IT Engineering, École de technologie supérieure, Montréal, QC, CanadaNetwork Technology Lab, Department of Electrical Engineering, École de technologie supérieure, Montréal, QC, CanadaSpan Loss is a pivotal characteristic of optical networks, and its accurate prediction enables adjustment for optimal performance and proactive monitoring. Deep learning models such as transformers, with their self-attention mechanism, have shown potential for various prediction tasks. In this study, we propose the Transformer-XL (Extra Long) model for single-step and multi-step forecasting, trained with field data. We report on models predicting span loss from 15 minutes to 5 days, using window sizes of 15 minutes to 10 days. The single-step model's average Absolute Maximum Error (AME) is better than the naive model by 2.13 dB and outperforms linear regression by 0.05–0.32 dB across different window sizes. Our single-step model also achieves better performance than the Recurrent Neural Network (RNN) with an AME improvement of 0.02 dB. The average AME of our multi-step model exceeds the naive model's performance by a range of 2.95-3.05 dB, linear regression by a substantial 0.02-0.15 dB and RNN by a range of 0.04-0.54 dB across different window sizes and forecast horizons. Based on Root Mean Square Error (RMSE), the single-step model performs better than the naive approach across various window sizes by 0.07 dB, achieves up to 0.07 dB improvement over linear regression, and delivers comparable results to RNN. Moreover, our multi-step model improves upon the naive approach with RMSE by 0.04 dB and RNN by 0.02 across various window sizes and forecast horizons. It also demonstrates a slight improvement over linear regression.https://ieeexplore.ieee.org/document/11006412/Attention mechanismdeep learning modelsmulti-step predictionoptical networkssingle-step predictionspan loss
spellingShingle Maryam Hedayatnejad
Yinqing Pei
David Boertjes
Dacian Demeter
Christian Desrosiers
Christine Tremblay
Multi-Step Span Loss Prediction in Optical Networks Using Multi-Head Attention Transformers
IEEE Photonics Journal
Attention mechanism
deep learning models
multi-step prediction
optical networks
single-step prediction
span loss
title Multi-Step Span Loss Prediction in Optical Networks Using Multi-Head Attention Transformers
title_full Multi-Step Span Loss Prediction in Optical Networks Using Multi-Head Attention Transformers
title_fullStr Multi-Step Span Loss Prediction in Optical Networks Using Multi-Head Attention Transformers
title_full_unstemmed Multi-Step Span Loss Prediction in Optical Networks Using Multi-Head Attention Transformers
title_short Multi-Step Span Loss Prediction in Optical Networks Using Multi-Head Attention Transformers
title_sort multi step span loss prediction in optical networks using multi head attention transformers
topic Attention mechanism
deep learning models
multi-step prediction
optical networks
single-step prediction
span loss
url https://ieeexplore.ieee.org/document/11006412/
work_keys_str_mv AT maryamhedayatnejad multistepspanlosspredictioninopticalnetworksusingmultiheadattentiontransformers
AT yinqingpei multistepspanlosspredictioninopticalnetworksusingmultiheadattentiontransformers
AT davidboertjes multistepspanlosspredictioninopticalnetworksusingmultiheadattentiontransformers
AT daciandemeter multistepspanlosspredictioninopticalnetworksusingmultiheadattentiontransformers
AT christiandesrosiers multistepspanlosspredictioninopticalnetworksusingmultiheadattentiontransformers
AT christinetremblay multistepspanlosspredictioninopticalnetworksusingmultiheadattentiontransformers