Efficient Classification of Polarization Events Based on Field Measurements

摘要

We present rare-event classification of polarization transients based on field measurements with data augmentation combined with robot-generated fiber-disturbance data. We compare machine learning methods for accuracy and required number of training sample traces.