Enhancing nonlinear compensation efficiency with multi-task neural networks for coherent optical systems

Area of Science:

• Optical communication systems engineering
• Computational intelligence in telecommunications
• Signal processing for optical networks

Background:

• Traditional nonlinear compensation methods in high-speed optical systems suffer from high computational complexity.
• Existing techniques struggle to balance compensation accuracy with computational efficiency.

Purpose of the Study:

• To develop a low-complexity nonlinear compensation method for coherent optical communication systems.
• To reduce computational overhead without sacrificing system performance.

Main Methods:

• A multi-task neural network (MT-NN) framework is proposed, leveraging shared network weights for simultaneous symbol processing.
• Complexity-aware mean square error (MSE) and partial grid search optimization are employed.
• Transfer learning (TL) is integrated to improve training efficiency.

Main Results:

• The MT-NN approach significantly lowers computational complexity across various optical transmission scenarios.
• Compensation accuracy is maintained, demonstrating effectiveness without performance compromise.
• Experimental results show a superior accuracy-efficiency trade-off compared to single-task neural networks (ST-NN).

Conclusions:

• The proposed MT-NN-based nonlinear compensation offers a practical and efficient solution for next-generation optical communication systems.
• This method addresses the challenge of computational complexity in high-speed optical networks.
• The study highlights the potential of multi-task learning and transfer learning for advanced optical communication signal processing.