We are pleased to announce our contribution presented at OFC 2025:
Sasipim Srivallapanondh, Pedro Freire, Giuseppe Parisi, Mariano Devigili, Nelson Costa, Bernhard Spinnler, Antonio Napoli, Jaroslaw E. Prilepsky, and Sergei K. Turitsyn,
“Weight-Clustered Neural Networks for Low-Complexity Nonlinear Equalization in Digital Subcarrier Multiplexing Systems” (pp. 1-3).
🔗 Access the paper: https://doi.org/10.5281/zenodo.16778825
Highlights:
- Proposes a weight clustering technique to compress neural network structures used for nonlinear equalization in DSCM systems.
- Achieves approximately 91% reduction in complexity relative to perturbation-based NN models while preserving strong optical performance.
- Demonstrates the potential of model compression for deploying real-time, efficient, AI-based equalizers in high-speed optical networks.