5.1SPFeb 25, 2022
Domain Adaptation: the Key Enabler of Neural Network Equalizers in Coherent Optical SystemsPedro J. Freire, Bernhard Spinnler, Daniel Abode et al.
We introduce the domain adaptation and randomization approach for calibrating neural network-based equalizers for real transmissions, using synthetic data. The approach renders up to 99\% training process reduction, which we demonstrate in three experimental setups.