InterDigital, a mobile, video and AI technology research and development company, has announced its participation in the 2024 Brooklyn 6G Summit, taking place from October 23-25. At the show, InterDigital and researchers from The University of Texas at Austin will jointly display the latest solutions to validate AI solutions for wireless networks and shape how AI will be used for 5G and 6G.
Jointly organized by Nokia and the NYU Wireless Research Center, the Brooklyn Summit is an annual gathering of communications industry leaders contributing to the future of wireless technology. The theme of this year’s discussions is “6G – From vision to action.”
A key challenge in applying AI/ML to wireless networks is managing model training overhead and ensuring user data privacy. To address these issues, InterDigital and The University of Texas at Austin will showcase our research on Over-the-Air Federated Learning (OTA-FL) by training a real-time image classification in a live demo.
While traditional federated learning (FL) requires each device to send its model updates using separate dedicated radio resources to a central server for aggregation, in OTA-FL, the aggregation of model updates happens directly "over-the-air". The joint demo between InterDigital and The University of Texas at Austin will use software defined radios (SDRs) to simultaneously transmit model updates, displaying real-time OTA-FL training. The simultaneous transmission of model updates reduces overhead and improves energy efficiency, especially as the number of users increases.
“As industry prepares for the increased use of AI, new solutions are required to enable AI – especially for the training of AI models,” said Milind Kulkarni, VP and Head of Wireless Labs at InterDigital. “Our expert teams at InterDigital are leading critical research in AI and its enablement in wireless networks, and this work is one example of the many novel approaches being considered.”
“OTA-FL offers an incredible opportunity to have model weight updates from large number of devices without proportional increase in precious spectrum and time resources. It is a whole new paradigm of training where the wireless medium computes for us,” said UT-Austin Professor Kaushik Chowdhury. “Through this collaboration and support from InterDigital, as well as from the US National Science Foundation, we are excited to be among the first demonstrators of OTA-FL on real-world experimental testbeds.”
In addition to their AI demo showcase, InterDigital’s Senior Principal Engineer, Phil Pietraski, will participate in a panel discussion on “Energy Efficiency in AI/ML Networks,” alongside speakers from NVIDIA, Fraunhofer Heinrich Hertz Institute (HHI), Rohde & Schwarz, and MediaTek. The discussion will take place on Friday, October 25 at 1:40pm ET in the NYU Wireless Pfizer Auditorium. Click here to learn more about this panel discussion on AI/ML for wireless networks.
Click here to learn more about the 2024 Brooklyn 6G Summit.
