Anomaly Detection and Localization in Optical Networks Using Vision Transformer and SOP Monitoring
Khouloud Abdelli, Matteo Lonardi, J. Gripp, Daniela Gallon Corrêa, Samuel L. I. Olsson, Fabien Boitier, Patricia Layec
Abstract
We introduce an innovative vision transformer approach to identify and precisely locate high-risk events, including fiber cut precursors, in state-of-polarization derived spectrograms. Our method achieves impressive 97% diagnostic accuracy and precise temporal localization (6-ms- RMSE).
Topics & Concepts
TransformerComputer scienceArtificial intelligenceAnomaly detectionSpectrogramComputer visionPattern recognition (psychology)EngineeringVoltageElectrical engineeringOptical Network TechnologiesSpectroscopy Techniques in Biomedical and Chemical ResearchAnomaly Detection Techniques and Applications