17 Mar 2026
14:00 - 16:00
Room 502A
This panel addresses the critical need for an effective selection framework for machine learning (ML) algorithms in optical communication networks. ML shows great potential in challenging tasks such as nonlinearity compensation, network performance monitoring, and fault management. The plethora of state-of-the-art ML algorithms makes choosing the best fit difficult. What data is used for training and evaluation is equally important. Bringing together researchers, industry experts, and network operators, the panel will review current ML applications, highlighting benefits and challenges.
The key points in this panel will be:
- The discussion will focus on suitability and limitations of different ML algorithms for key optical communication tasks, aiming to establish selection criteria for dataset quality, performance, interpretability, efficiency, and scalability.
- The scope will include ML-based signal processing, transmission data analysis, network optimization, and trends such as photonic neural networks and standardized ML models.
- The goal is to provide practical guidelines and foster collaboration, ultimately advancing intelligent, ML-driven optical networks towards real-world demands.
Organizers
-
Boris Karanov
Karlsruher Institut für Technologie, Germany
-
Lakshmi Narasimhan
Indian Institute of Technology, India
-
Deepa Venkitesh
Indian Institute of Technology Madras, India
Panelists
-
Steve Hranilovic
McMaster University , Canada
Optical wireless channels differ greatly from their analogous guided fiber optical counterparts. This difference arises due to their inherently different channel models; due shadowing and diffuse reflections (indoors over short distances) as well as due to atmospheric impacts (over longer-range links). In this talk, a brief overview of the unique properties, constraints and applications of optical wireless channels will be provided some ML approaches that have been taken to address them.
Steve Hranilovic (S’94-M’03-SM’07-F’22) received the B.A.Sc. degree with honours in electrical engineering from the University of Waterloo, Canada in 1997 and M.A.Sc. and Ph.D. degrees in electrical engineering from the University of Toronto, Canada in 1999 and 2003 respectively.
He is a Professor in the Department of Electrical and Computer Engineering, McMaster University (Hamilton, Ontario, Canada) where he currently serves as the Vice-Provost and Dean of Graduate Studies. During 2010-2011 he spent his research leave as Senior Member, Technical Staff in Advanced Technology for Research in Motion, Waterloo, Canada. His research interests are in the areas of free-space and optical wireless communications, digital communication algorithms, and electronic and photonic implementation of coding and communication algorithms. He is the author of the book Wireless Optical Communication Systems (New York:Springer, 2004). Dr. Hranilovic is a Fellow of the IEEE and of Optica, a Fellow of the Canadian Academy of Engineering and is a licensed Professional Engineer in the Province of Ontario. In 2016 the title of University Scholar was conferred upon him by McMaster University. He has served as an Associate Editor for the Journal of Optical Communications and Networking and an Editor for the IEEE Transactions on Communications in the area of Optical Wireless Communications.
-
Mathieu Chagnon
Arycs Technologies, United States
Machine Learning algorithms can be applied in different segments of optical communication systems. We review which segments are better suited to leverage ML and discuss some of the key benefits and challenges of this technology.
Mathieu Chagnon obtained his Ph.D. from McGill University, Canada, from the Department of Electrical and Computer Engineering in the Photonics Systems Group, during which he was awarded several Fellowships and Scholarships. He joined Nokia Bell Labs as a Member of Technical Staff and Researcher Engineer in the Smart Optical Fabric, Devices, Transmission, and DSP Lab, in Stuttgart, Germany. Dr. Chagnon subsequently joined Infinera, USA, as a Principal Coherent Optical System Architect, in the Advanced Optical System Group. In 2022, Dr. Chagnon joined Arycs Technologies as Director of Optical Systems Engineering. Dr. Chagnon co-authored a book chapter on high-speed interconnects for data center networking and gave Invited and Tutorial talks at all major optical communication conference venues (OFC, ECOC, and the Asia Communications and Photonics Conference). He has authored/co-authored a substantial body of peer-reviewed journals and proceedings including several post deadlines and patents.
-
Tim O'Shea
DeepSig AI, United States
The physical layer is transforming rapidly, leveraging AI/ML methods to improve core signal processing functions and to introduce new capabilities and services. ML methods have offered a path to more efficient communications, pushing spectral efficiency and resilience of our communications systems further than previously possible by leveraging real world channel and measurement data. They have also helped to make wireless systems more aware with a high-level semantic understanding of activity in the radio frequency spectrum – allowing for deep analytics, intelligent automation, and improved network operations and security capabilities. This talk will highlight the key enablers, techniques, and path from early research to mature productized solutions ready for deployment. We will highlight the work at DeepSig on this topic, the critical intersections with OpenRAN, open source, 6G systems, and defense applications – and will highlight how DeepSig is using ML in AI-Native RAN within the recently launched Linux foundation’s OCUDU project driving open and intelligent AI-Native RAN platforms on top of the “Linux of RAN”.
Dr. Tim O'Shea is the CTO at DeepSig Inc, where he is focused on enhancing wireless performance using AI/ML in the physical layer where they have fielded the world's first AI-Native OpenRAN network and are actively involved in 3GPP 6G and AI-Native Air Interface standardization and prototyping. He also serves on the FCC TAC, for Virginia Tech Faculty, AI-RAN Alliance TSC, OCUDU TSC, and with IEEE's Machine Learning for Communications initiatives, and is the inventor on over 100 patent filings and publication. He has previously worked with Hawkeye 360, Federated Wireless, GNU Radio, Cisco Systems, and the Department of Defense.
-
Tingjun Chen
Duke University, United States
Fiber optical networks have been widely deployed at different scales to form the core infrastructure of today’s Internet backbone, telecommunication networks, and smart connected communities. These networks can deliver high bandwidth data services at deterministic low latency leveraging advanced techniques and hardware such as wavelength-division multiplexing (WDM) technique and reconfigurable add-drop multiplexer (ROADM) units. Accurate modeling and performance estimation, as well as autonomous adaptation and reconfiguration of optical links are essential for optical system designs, particularly due to wavelength-dependent gain spectrum of optical amplifiers and fiber nonlinearity. Moreover, each fiber-optic cable can also serve as a high-resolution sensor since it is sensitive to different environmental effects (e.g., vibration and temperature) due to the linear and nonlinear light scattering. In this talk, I will first present the modeling of erbium-doped fiber amplifiers (EDFAs) using machine learning (ML) and its application in signal quality estimation in multi-span ROADM systems. I will then present our investigation on the coexistence of heterogeneous communication, fiber sensing, and radio-over-fiber signals that co-propagate on the same fiber, focusing on their impact on key performance metrics including bit error rate (BER) and sensing resolution. Lastly, I will highlight a series of measurements and field trials that support our research, conducted on the PAWR COSMOS platform in Manhattan, NYC, and the Duke BlueFrog testbed in the Research Triangle, NC.
Tingjun Chen is the Nortel Networks Assistant Professor of Electrical & Computer Engineering at Duke University, with a secondary appointment in Computer Science. His research focuses on networking, communication, sensing, and energy-efficient computing for wireless, mobile, and optical networked systems, bridging theoretical foundations and experimental platforms. He received his Ph.D. degree in Electrical Engineering from Columbia University in 2020, and his B.Eng. degree in Electronic Engineering from Tsinghua University in 2014. Between 2020–2021, he was a Postdoctoral Associate at Yale University. He has received multiple awards, including the NSF CAREER Award, Google Research Scholar Award, IBM Academic Awards, NVIDIA Academic Awards, Columbia Engineering Morton B. Friedman Memorial Prize for Excellence, Columbia University Eli Jury Award, and the Facebook Fellowship. He is also a co-recipient of several paper awards from ACM CoNEXT, ACM MobiHoc, IEEE MTT-S IMS, IEEE/Optica OFC, and ECOC. His Ph.D. thesis was recognized by the ACM SIGMOBILE Dissertation Award Runner-up.
-
Toshiaki Akino
Mitsubishi Electric Research Laboratories, United States
Over the past decade, research at the intersection of AI and optics has grown exponentially, paralleling the broader AI revolution. Scaling laws in AI show that performance improves with more data, larger models, and increased compute. Optical systems naturally fit this paradigm, generating terabits-grade streams of rich physical-layer data—an abundant resource for training data-driven algorithms. Yet the relationship is not unidirectional. While AI is transforming photonic systems, optics may also help address AI’s growing energy and scalability challenges. Photonic computing offers ultrafast, massively parallel, high data-rate, and potentially low-energy processing beyond conventional electronic CPU/GPU/TPU. This talk argues for a shift in perspective: from applying AI to optical systems to leveraging optical physics to shape the future of AI. The key question may not be which algorithms to use—but which physics will allow them to scale.
Toshiaki Koike-Akino received the B.S. degree in electrical and electronics engineering, M.S. and Ph.D. degrees in communications and computer engineering from Kyoto University, Kyoto, Japan, in 2002, 2003, and 2005, respectively. During 2006–2010 he was a Postdoctoral Researcher at Harvard University, and is currently a Distinguished Research Scientist at Mitsubishi Electric Research Laboratories (MERL), Cambridge, MA, USA. His research interests include signal processing, communications, device, sensing, quantum, and artificial intelligence. He received the 2008 Ericsson Young Scientist Award, the IEEE GLOBECOM’08 Best Paper Award, and the IEEE GLOBECOM’09 Best Paper Award. He is a Fellow of Optica.