Failure management in optical networks with ML: a tutorial on applications, challenges, and pitfalls [Invited]

Abstract

This tutorial identifies and discusses the main design choices and challenges arising in the application of machine learning (ML) to optical network failure management (ONFM), including quality of transmission estimation, failure detection, prediction, root-cause identification, localization, and magnitude estimation. We focus on input data preparation and on interpreting and validating model outputs, tackling data scarcity, data confidentiality, model explainability, uncertainty quantification, and other critical factors, in order to highlight the potential risks for practitioners when adopting ML-based solutions for ONFM. An overview of publicly available datasets is also provided.