Migrating the Intelligence from the Cloud to Users – Mobile AI from Federated Learning to On-device Training

  • 340


With the rapid advancements in mobile sensing and computing technology, there is a surge of innovative intelligent applications that monitor and manage our daily lives ubiquitously. Traditional mobile sensing/computing applications used to collect user information locally and transmit data to a more powerful server for intelligent data processing operations. However, more recent developments showcase a shift towards a more localized and personalized approach. This involves processing data collected locally on user devices, enhancing efficiency, and prioritizing privacy. This presentation will delve into the applications driving this paradigm shift, bringing intelligence from the cloud to user devices. Additionally, we will explore novel machine learning frameworks supporting localized intelligence experiences. This includes introducing mechanisms to reduce model complexity, enabling memory-efficient on-device model training, and facilitating efficient federated learning.


JeongGil Ko is an associate professor in the School of Integrated Technology, College of Computing and the Department of Biomedical Systems Informatics, College of Medicine, and serves as the Associate Vice President for Information and Communications at Yonsei University. He is also affiliated with the Pohang University of Science and Technology (POSTECH) as an adjunct professor at the Graduate School of Artificial Intelligence. JeongGil received his B.Eng. in Computer Science and Engineering from Korea University (2007) and received his Ph.D. in Computer Science from the Johns Hopkins University (2012). He was at the Department of Computer Science, Stanford University as a visiting researcher (2010) and was a senior researcher at the Electronics and Telecommunications Research Institute (2012-2015). Prior to joining Yonsei, JeongGil was an assistant professor at the Department of Software and Computer Engineering at Ajou University (2015-2019). He is a recipient of the Abel Wolman Fellowship awarded by the Whiting School of Engineering at the Johns Hopkins University and a senior member of the ACM and IEEE. JeongGil has served on the program committee for many top conferences in the mobile and ubiquitous computing field (ACM MobiCom, MobiSys, SenSys, IEEE PerCom in particular), was the program chair for ACM UbiComp 2022, is an associate editor for renown academic journals including PACM IMWUT and IEEE TMC, and will serve as the general co-chair for ACM MobiSys 2024. His research interests are in the general area of developing mobile/embedded sensing systems with ambient intelligence and systems-related topics for enabling efficient XR applications.