Domain adaptation: The key enabler of neural network equalizers in coherent optical systems
Optical Fiber Communication Conference, 2022•opg.optica.org
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.
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.
Abstract
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.
opg.optica.org