Active Learning with Gaussian Process Regression and Physical Models for Robust SNR Estimation

Xiaoyan Ye; Mariane  Mansour; Md Saifuddin Faruk; Charles Laperle; Michael Reimer; Maurice O'Sullivan; Seb J. Savory

Active Learning with Gaussian Process Regression and Physical Models for Robust SNR Estimation

Xiaoyan Ye
, Mariane Mansour
, Md Saifuddin Faruk
, Charles Laperle
, Michael Reimer
, Maurice O'Sullivan
, Seb J. Savory

School of Computer Science & Engineering

University of Cambridge
Ciena, ottawa, Ontario, Canada
Ciena Canada, Inc. Ottawa

Research output: Contribution to conference › Paper › peer-review

281 Downloads (Pure)

Abstract

We demonstrate improved performance using active learning for both GPR
and hybrid models to predict SNR using experimental data from a 15-channel WDM
system over 1000km. Physical model interpreted GPR agrees with interpreting
measured data

Original language	English
Publication status	Published - 2025

Access to Document

OFC_2025_SNR estimationAccepted author manuscript, 402 KB

Cite this

@conference{b84cbc391e3c458ba921b31f885f67ff,

title = "Active Learning with Gaussian Process Regression and Physical Models for Robust SNR Estimation",

abstract = "We demonstrate improved performance using active learning for both GPR and hybrid models to predict SNR using experimental data from a 15-channel WDM system over 1000km. Physical model interpreted GPR agrees with interpreting measured data",

author = "Xiaoyan Ye and Mariane Mansour and Faruk, \{Md Saifuddin\} and Charles Laperle and Michael Reimer and Maurice O'Sullivan and Savory, \{Seb J.\}",

year = "2025",

language = "English",

}

TY - CONF

T1 - Active Learning with Gaussian Process Regression and Physical Models for Robust SNR Estimation

AU - Ye, Xiaoyan

AU - Mansour, Mariane

AU - Faruk, Md Saifuddin

AU - Laperle, Charles

AU - Reimer, Michael

AU - O'Sullivan, Maurice

AU - Savory, Seb J.

PY - 2025

Y1 - 2025

N2 - We demonstrate improved performance using active learning for both GPR and hybrid models to predict SNR using experimental data from a 15-channel WDM system over 1000km. Physical model interpreted GPR agrees with interpreting measured data

AB - We demonstrate improved performance using active learning for both GPR and hybrid models to predict SNR using experimental data from a 15-channel WDM system over 1000km. Physical model interpreted GPR agrees with interpreting measured data

M3 - Paper

ER -

Active Learning with Gaussian Process Regression and Physical Models for Robust SNR Estimation

Abstract

Access to Document

Fingerprint

Cite this