Optical fibre networks underpin the global communications infrastructure, carrying >95% of all digital data. These optical networks operate on different time- and distance-scales – spanning long-distances (trans-oceanic, inter-satellite) and inter-city, as well as interconnecting servers and GPUs within data centres. The challenge for next-generation networks is in making them intelligent and adaptable to changes in demand and applications in the future.
Advancing AI and scientific research requires optimizing optical networks from the component through to the physical network layer. The physical topology and nonlinear properties of the optical fibre form a key bottleneck affecting achievable throughput in these networks. Optimising the throughput involves solving two NP-hard combinatorial optimisation problems: (i) the physical topology design problem and (ii) the routing and wavelength assignment problem by including the physical properties in ultrawideband optical network design – through and for artificial intelligence.
This talk will consider both of these challenges and their influence the maximum achievable throughput – a vital objective for the future scaling of AI algorithms.
References:
R. Matzner, A. Ahuja, R. Sadeghi, M. Doherty, S. J. Savory, and P. Bayvel, ‘Topology Bench: Systematic Graph Based Benchmarking for Core Optical Networks’, arXiv preprint arXiv:submit/5982227 (2024).
R. Matzner, D. Semrau, R. Luo, G. Zervas, and P. Bayvel, ‘Making intelligent topology design choices: understanding structural and physical property performance implications in optical networks ’, Journal of Optical Communications and Networking, vol. 13, no. 8, p. D53, Aug. 2021.
R. Matzner, R. Luo, G. Zervas, and P. Bayvel, ‘Expanding Graph Neural Networks for Ultra-Fast Optical Core Network Throughput Prediction to Large Node Scales’, in European Conference on Optical Communication (ECOC 2022), paper We5.61, Optica Publishing Group, Sep. 2022, p. We5.61.
R. Matzner, R. Luo, G. Zervas, and P. Bayvel, ‘Ultra-fast Optical Network Throughput Prediction using Graph Neural Networks’, in 26th International Conference on Optical Network Design and Modelling, May 2022, p. 3.
R. Matzner, R. Luo, G. Zervas, and P. Bayvel, ‘Intelligent performance inference: A graph neural network approach to modelling maximum achievable throughput in optical networks’, APL Machine Learning, vol. 1, no. 2, p. 026112, May 2023.
Distinguished Speaker: Prof. Polina Bayvel
[]
Professor Polina Bayvel is the Head of the (https://www.ee.ucl.ac.uk/ong/) (Department of Electronic & Electrical Engineering), UCL, a research group she set up in 1994. Her research interests are in the area of optical communications and networks on different time- and length scales (from trans-oceanic distances to intra- and inter-data centre communications and routing. This includes the analysis and implementation of intelligent optical networks, wavelength routing, high-speed optical transmission, and the study and mitigation of optical fibre nonlinearities. Following her PhD, she worked as a systems engineer in the STC Submarine Systems, Greenwich (now Alcatel Submarine Networks) on the first amplified optical transmission and in Nortel (Harlow Labs) on optical network planning.
In 1994 she received a Royal Society University Research fellowship and moved to UCL to set up ONG, the first academic systems engineering group in optical networks.
She is a Fellow of the Royal Society, Royal Academy of Engineering, IEEE and Optica. She is the 2021 recipient (and first woman) of the Thomas Young Medal of the UK Institute of Physics and the first woman to be awarded the 2023 Royal Society Rumford Medal since it was introduced in 1800! In 2024 she received the Humboldt Foundation Research Prize.
Author/co-author of more than 500 refereed journal and conference papers, she has led the UK EPSRC Programme Grant (https://www.unloc.net/) (2012-2018), focused on unlocking – and maximising – the capacity of optical fibre communications and currently leads the EPSRC Programme Grant TRANSNET (2018-2026): https://transnet.org.uk/ focused on transforming optical networks for the cloud.
Agenda:
11:00 – 11:30 התכנסות וכיבוד קל
11:30 – 12:30 הרצאה של פרופ' פולינה בייבל
Room: Lecture Hall 001, Bldg: Broadcom Semiconductors, Tel Aviv University, Dr. George Wise St. 17, Tel Aviv, Tel Aviv District, Israel, 69978