Allegro EU Project Publishes Low-Complexity Deep Learning Method for Optical Network Failure Management

We’re proud to highlight our recent publication at ONDM 2024:

Lareb Zar Khan, João Pedro, Omran Ayoub, Nelson Costa, Andrea Sgambelluri, Lorenzo De Marinis, Antonio Napoli, and Nicola Sambo,
“Optimizing Deep Learning–Based Failure Management in Optical Networks by Monitoring Relative Neural Activity,” ONDM 2024, pp. 1–3.
🔗 Access the paper: https://doi.org/10.23919/ONDM61578.2024.10582707

Highlights:

• Introduces a novel technique that prunes neural networks based on neuron activity—preserving only the most important neurons (“iterative neural removal”).
• Achieves dramatic efficiency gains: up to 96% reduction in computational complexity and 87% less memory usage, with minimal impact on predictive performance.
• Addresses a critical need for deploying deep learning models in real-world optical network operations with computational constraints.

This publication demonstrates the Allegro EU Project’s commitment to practical, scalable artificial intelligence solutions for next-generation optical infrastructure.