Data augmentation to improve performance of neural networks for failure management in optical networ

Data augmentation to improve performance of neural networks for failure management in optical networ

Abstract:

A variational-autoencoder based data augmentation technique was investigated to improve the quality and increase the amount of data for optical network failure management. Augmentation provided significant performance improvement in terms of reduction in machine-learning training time for soft-failure detection (37.56%) and cause identification (66.5%).