A Three-Stage Training Framework for Customizing Link Models for Optical Networks

Abstract

We propose a link model customization framework to increase modeling accuracy for each specific link in an optical network. In addition, an active acquisition method is employed in this framework to improve tolerance to link parameter uncertainties.