Machine Learning for optical spectrum analysis

In this chapter, we propose Machine Learning (ML) based solutions for Optical Spectrum Analysis of Elastic Optical Networks (EON). Specifically, we present novel failure detection and identification solutions utilizing the optical spectrum traces captured by cost-effective coarse-granular Optical Spectrum Analyzers (OSA). We demonstrate the effectiveness of the developed solutions for detecting and identifying filter-related failures in the context of Spectrum-Switched Optical Networks (SSON), as well as transmitter-related laser failures in Filterless Optical Networks (FON). Such detection and identification can contribute to the cost reduction and lowering the required margin in optical networks.