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

This addresses the need for more adaptable and efficient signal processing in optical communication systems, representing an incremental advancement by applying multi-task learning to a specific domain.

The paper tackled the problem of improving the flexibility of neural network-based equalizers in coherent optical systems by proposing multi-task learning for the first time, resulting in a single equalizer that enhances Q-factor by up to 4 dB compared to conventional digital compensation without needing retraining across variations in launch power, symbol rate, or transmission distance.