Multi-Task Learning to Enhance Generalizability of Neural Network Equalizers in Coherent Optical Systems

TL;DR

This paper introduces a multi-task learning approach to neural network equalizers in coherent optical systems, significantly enhancing their adaptability and performance across various transmission conditions without re-training.

Contribution

It presents the first application of multi-task learning to neural network equalizers, improving their generalizability and performance in coherent optical communications.

Findings

Q-factor improved by up to 4 dB over CDC

Equalizer maintains performance across different launch powers

No re-training needed for different system variations

Abstract

For the first time, multi-task learning is proposed to improve the flexibility of NN-based equalizers in coherent systems. A "single" NN-based equalizer improves Q-factor by up to 4 dB compared to CDC, without re-training, even with variations in launch power, symbol rate, or transmission distance.