Machine learning in optical communication systems
Thèmes scientifiques :
- D - Télécommunications : compression, protection, transmission
Nous vous rappelons que, afin de garantir l'accès de tous les inscrits aux salles de réunion, l'inscription aux réunions est gratuite mais obligatoire.
10 personnes membres du GdR ISIS, et 22 personnes non membres du GdR, sont inscrits à cette réunion.
Capacité de la salle : 300 personnes.
Thursday, March 25, 2021, 2pm - 5:30pm
In this workshop, we are interested in covering an overview of machine learning (ML) algorithms and applications for optical communication systems and networks. ML is based on an idea that synthetic or real data can be used to train systems in order to enable them to make decisions or predictions on new unknown data. Most ML algorithms deal with two tasks: regression and classification (or clustering). Although ML is not a new field, recent important increases in computational power and access to abundant quantities of data contributed to the advent of novel ML methods applied in several fields.
Researchers in the field of optical communications are no strangers to regression and classification problems tackled with probability theory and an understanding of the problem's underlying physics. However, as the sources of transmission impairments are becoming more numerous and complex for high-rate links and networks, explicit characterizations of these impairments, their mitigation and the prediction of their impact on the quality of transmission become hard to analyze. As a consequence, applications of ML techniques range from compensation of fiber non linearity and transceiver imperfections to optical performance monitoring and software-defined networking. Indeed, besides the developments related to the physical layer, optical network architectures and operations are undergoing an important evolution towards adaptive provisioning of resources and fast discovery of faults to minimize system outage.
Extracting patterns from numerous physical-layer and network-layer parameters naturally calls for ML algorithms. Through a coverage of ML applications in optical communications and networking, we hope to provide through this workshop a better understanding of the most suitable use cases of ML where it can play a unique role.
Call for contributions
If you are willing to contribute to the workshop by giving a talk related to one of the aforementioned ML applications to optical fiber transmissions, please contact the organizers before March 18, 2021. Participation of PhD students and postdoctoral researchers is strongly encouraged.
Program (contributions accepted until March 18, 2021)
In each slot, 10 minutes are allocated for Q&A.
Introduction - Catherine Lepers & Elie Awwad
Machine learning-aided quality of transmission(QoT) estimation -
Yvan Pointurier, Huawei France
A novel data augmentation technique to reduce the complexity of receiver-based DSP in optical telecommunications -
Vladislav Neskorniuk, Aston Institute of Photonic Technologies
The Potential of Artificial Intelligence (AI) in Optical Transport Network -
Ahmed Triki, Orange Labs
Design of Submarine Optical Fiber Links Optimized through Reinforcement Learning -
Maria Ionescu, Nokia Bell Labs France
Efficient equalization in nonlinear fiber-optic communications using a convolutional recurrent neural network -
Abtin Shahkarami, Télécom Paris
|5:10-5:20||Conclusion - Catherine Lepers & Elie Awwad|
A workshop organized by Digicosme and GDR-ISIS.
Elie Awwad (email@example.com)
Catherine Lepers (firstname.lastname@example.org)