Deep Learning-Based Classification for QoT Estimation in SMF and FMF Links

Prediction of the quality of transmission (QoT) in optical communication networks is a critical task to ensure reliable transmission and optimized operation. Accurate QoT estimation is essential for enabling future reconfigurable optical networks, where lightpaths are set up based on performance predictions. There are three main categories of QoT estimation methods: exact analytical models offer high accuracy but require significant computational power, making them unsuitable for real-time applications, approximate formulas are faster to compute but sacrifice some accuracy, machine learning (ML) and deep learning (DL) methods potentially offer high accuracy with lower computational demands, making them promising candidates for real-time scenarios. DL has emerged as a powerful tool for modeling complex nonlinear relationships and has shown promise in QoT estimation. In this paper, we propose a DL-based classifier to predict whether a lightpath meets the required system performance, both in single-mode fiber (SMF) and few-mode fiber (FMF) links. Using synthetic datasets generated from the enhanced Gaussian noise (EGN) model, we compare DL-based classifier with ML-based classifiers, DL-based regressor, and closed-form models (CF-GN and CF-EGN). Results show that the proposed DL-based method outperforms the conventional ML-based method and closed-form models in both SMF and FMF links in terms of various classification metrics. Moreover the DL-based classifier outperforms the DL-based regressor in terms of complexity, while achieving nearly the same performance in both SMF and FMF links. This superiority stems from DL's capabilities: DL models can effectively capture the intricate interactions between various factors that influence QoT, DL networks can learn layered representations of the data, allowing for a deeper understanding of the underlying patterns, and DL models automatically extract relevant features from the data, eliminating the need for manual feature engineering. This work demonstrates the potential of DL for real-time and accurate QoT estimation in optical networks. The proposed DL-based classifier can be used to optimize link configurations, predict network performance, and enhance overall network reliability.