Multi-Task Learning to Enhance Generalizability of Neural Network Equalizers in Coherent Optical Systems
Published in European Conference on Optical Communication (ECOC 2023), accepted preprint, 2023
Multi-task learning is proposed to enhance the flexibility of neural network-based equalizers in coherent optical systems. A single NN model improves Q-factor by up to 4 dB compared to CDC without requiring re-training, even under varying launch power, symbol rate, or transmission distance.
@inproceedings{srivallapanondh2023multi,
title={Multi-task learning to enhance generalizability of neural network equalizers in coherent optical systems},
author={Srivallapanondh, Sasipim and Freire, Pedro J and Alam, Ashraful and Costa, Nelson and Spinnler, Bernhard and Napoli, Antonio and Sedov, Egor and Turitsyn, Sergei K and Prilepsky, Jaroslaw E},
booktitle={IET Conference Proceedings CP839},
volume={2023},
number={34},
pages={640--643},
year={2023},
organization={IET}
}