TRANSNET speakers at ECOC

The European Conference on Optical Communication (ECOC) is being held virtually in Brussels from 6-10 December, and several members of the TRANSNET Programme will be speaking, including our PI Polina Bayvel.

ECOC was originally scheduled for September, set to return to Brussels 25 years after it was first hosted there in 1995, but the Covid-19 pandemic pushed back the conference to December – it will still be hosted in Belgium... but the programme will be entirely online.

ECOC is the leading European conference in the field of optical communication, and although attendees cannot be there in person this year all the latest research and technological progress will continue to be presented in selected papers, plenary talks, workshops and much more. Members of TRANSNET will be there along with the following contributions.

Tuesday 8 December

Paper: Intelligent design of optical networks: which topology features help maximise throughput in the nonlinear regime? (invited)
Authors: Bayvel. Polina, Luo. Ruijie, Matzner. Robin, Semrau. Daniel and Zervas. Georgios
Session: Passive optical networks I (Tu1J)
Time: 08:00 and 18:00
Abstract: The overarching goal in intelligent network design is to deliver capacity when and where it is needed. The key to this is to understand which network topology characteristics impact the achievable network throughput. This is explored through the use of a new generative network model, taking into account physical network characteristics.

PaperEnd-to-End Learning in Optical Fiber Communications: Concept and Transceiver Design (invited)
Authors: Karanov. Boris, Bayvel. Polina and Schmalen. Laurent
Session: Advanced transmission concepts (Tu2F)
Time: 09:30 and 19:30
Abstract: We implement a complete fiber-optic communication system as an end-to-end computational graph using an artificial neural network (ANN)-based transceiver. We highlight transceivers implemented using feedforward or recurrent ANN, and illustrate their performance by an example.

Wednesday 9 December

Paper: Experimental Verification of Complex-Valued Artificial Neural Network for Nonlinear Equalization in Coherent Optical Communication Systems
Authors: Freire. Pedro J, Neskorniuk. Vladislav, Napoli. Antonio, Spinnler. Bernhard, Costa. Nelson, Prilepsky. Jaroslaw E, Riccardi. Emilio and Turitsyn. Sergei
Session: Machine learning for transceivers (We1D)
Time: 08:10 and 18:10
Abstract: We propose a novel design of neural network for mitigating the fiber nonlinearity, employing a structure based on physical modelling. The neural network achieved nearly 5 times BER reduction in a field trial when transmitting WDM 200G DP-16QAM over a 620 km legacy link.

Thursday 10 December

Paper: End-to-End Learning in Optical Fiber Communications: Experimental Demonstration and Future Trends (invited)
Authors: Karanov. Boris, Oliari. Vinicius, Chagnon. Mathieu, Liga. Gabriele, Alvarado. Alex, Aref. Vahid, Lavery. Domanic, Bayvel. Polina, and Schmalen. Laurent.
Session: End to end and supervised learning of optical channels (Th1D)
Time: 08:00 and 18:00
Abstract: Fiber-optic auto-encoders are demonstrated on an IM/DD test-bed, outperforming state-of-the-art signal processing. Algorithms for end-to-end optimization using experimentally collected data are discussed. The end-to-end learning framework is extended for performing optimization of the symbol distribution in probabilistically-shaped coherent systems

Paper: Simplifying the Supervised Learning of Kerr Nonlinearity Compensation Algorithms by Data Augmentation
Authors: Neskorniuk. Vladislav, Freire. Pedro J, Napoli. Antonio, Spinnler. Bernhard, Schairer. Wolfgang, Prilepsky. Jaroslaw E, Costa. Nelson and Turitsyn. Sergei
Session: End to end and supervised learning of optical channels (Th1D)
Time: 08:40 and 18:40
Abstract: We propose a data augmentation technique to improve performance and decrease complexity of the supervised learning of nonlinearity compensation algorithms. We demonstrate both numerically and experimentally that the augmentation allows reducing the training dataset size up to 6 times while keeping the same post-compensation bit-error rate.


We are sad to be missing the amazing beer and chocolate that Belgium is famous for, but happy to be sharing our latest work (virtually) with the ECOC community!