Optimizing Deep Learning-Based Failure Management in Optical Networks by Monitoring Relative Neural Activity

Despite demonstrating exceptional performance in optical networks, neural networks often receive criticism due to their significant computational complexity. To address this, we propose a novel relative-neural-activity-based algorithm to optimize neural networks for low computational complexity and memory footprint for failure management in optical networks, with failure identification as the considered use-case. The results suggest up to 96.31% and 87.12% reduction in computational and memory requirements, respectively.