Model-Centric versus Data-Centric Machine Learning for Soft-Failure Cause Identification in Optical Networks
Pedro, J. M.
; Costa, N.
Model-Centric versus Data-Centric Machine Learning for Soft-Failure Cause Identification in Optical Networks, Proc European Conference on Optical Communication ECOC, Glasgow, United Kingdom, Vol. , pp. - , October, 2023.
Digital Object Identifier:
Abstract
We compare model-centric and data-centric machine learning (ML) approaches to address the issue of insufficient training data for ML-based failure identification. The results suggest that a data-centric approach can improve classification accuracy by up to 7.1% on under-represented failures, albeit at a higher computational cost.
