Experimental validation of XPM mitigation using a generalizable multi-task learning neural network.

Area of Science:

• Optical Communications Engineering
• Machine Learning in Telecommunications
• Signal Processing for Optical Networks

Background:

• Long-haul dense wavelength-division multiplexing (DWDM) optical systems face signal impairments requiring advanced equalization.
• Traditional post-equalizers struggle with generalization and complexity in high-capacity optical transmission.
• Neural network (NN)-based equalizers offer potential but require robust generalization.

Purpose of the Study:

• To develop efficient neural network (NN)-based post-equalizers for coherent-detection DWDM systems.
• To enhance the generalization capability of NN equalizers using multi-task learning (MTL).
• To evaluate the performance and complexity of MTL equalizers against existing methods.

Main Methods:

• Implementation of a multi-task learning (MTL) framework for a shared neural network (NN) equalizer.
• Training and validation of the MTL equalizer model using experimental data from long-haul DWDM systems.
• Comparative analysis against single-task NN equalizer counterparts and previous approaches.

Main Results:

• The developed MTL equalizer model demonstrates superior performance and generalization compared to single-task NNs.
• Experimental data validates the effectiveness of MTL in improving equalizer performance.
• Significant reduction in equalizer complexity is achievable with MTL without compromising performance.

Conclusions:

• Multi-task learning (MTL) provides an effective strategy for enhancing NN-based post-equalizers in long-haul DWDM optical systems.
• MTL offers improved performance and generalization over single-task approaches.
• The proposed MTL equalizer achieves a favorable balance between performance and computational complexity.