Power Evolution Prediction and Optimization in a Multi-span System Based on Component-wise System Modeling
arXiv:2009.05348
Abstract
Cascades of a machine learning-based EDFA gain model trained on a single physical device and a fully differentiable stimulated Raman scattering fiber model are used to predict and optimize the power profile at the output of an experimental multi-span fully-loaded C-band optical communication system.