Differentiable Machine Learning-Based Modeling for Directly-Modulated Lasers

• Published in: IEEE Photonics Technology Letters
• Main Authors: Fernandez, Sergio, Hernandez, Jovanovic, Ognjen, Peucheret, Christophe, Ros, Francesco, Da, Zibar, Darko
• Other Authors: DTU Electrical Engineering Lyngby, Danmarks Tekniske Universitet = Technical University of Denmark (DTU), Institut des Fonctions Optiques pour les Technologies de l'informatiON (FOTON), Université de Rennes (UR)-Institut National des Sciences Appliquées - Rennes (INSA Rennes), Institut National des Sciences Appliquées (INSA)-Institut National des Sciences Appliquées (INSA)-École Nationale Supérieure des Sciences Appliquées et de Technologie (ENSSAT)-Centre National de la Recherche Scientifique (CNRS), 771878, ERC-CoG FRECOM Project, VIL29334, Villum YIP OPTIC-AI Project
• Format: Article in Journal/Newspaper
• Language: English
• Published: HAL CCSD 2024
• Subjects: Optical communication; machine learning; directly modulated laser; transformer; [SPI.OPTI]Engineering Sciences [physics]/Optics / Photonic; [INFO.INFO-LG]Computer Science [cs]/Machine Learning [cs.LG]; [INFO.INFO-MO]Computer Science [cs]/Modeling and Simulation; DML
• Online Access: https://hal.science/hal-04408658; https://hal.science/hal-04408658/document; https://hal.science/hal-04408658/file/Differentiable_Machine_Learning-Based_Modeling_for_Directly-Modulated_Lasers.pdf; https://doi.org/10.1109/lpt.2024.3350993

International audience End-to-end learning has become a popular method for joint transmitter and receiver optimization in optical communication systems. Such approach may require a differentiable channel model, thus hindering the optimization of links based on directly modulated lasers (DMLs). This is due to the DML behavior in the large-signal regime, for which no analytical solution is available. In this paper, this problem is addressed by developing and comparing differentiable machine learningbased surrogate models. The models are quantitatively assessed in terms of root mean square error and training/testing time. Once the models are trained, the surrogates are then tested in a numerical equalization setup, resembling a practical end-to-end scenario. Based on the numerical investigation conducted, the convolutional attention transformer is shown to outperform the other models considered.