Federated Learning in Heterogeneous Environments

Friday, August 6, 2021, 11:00 am - 12:00 pm PDTiCal
Virtual via Zoom WebinarThis event is open to the public.
AI Seminar
Jose Luis Ambite and Dimitri Stripelis, USC/ISI
Video Recording:

AI Seminar Series
Seminars for the Artificial Intelligence Division at USC Information Sciences Institute
YOU ONLY NEED TO REGISTER ONCE TO ATTEND THE ENTIRE SERIES- We will send you email announcements with details of the upcoming speakers.

Register in advance for this webinar:

After registering, you will receive a confirmation email containing information about joining the webinar.

There are situations where data relevant to machine learning problems are distributed across multiple locations that cannot share the data due to regulatory, competitiveness, or privacy reasons. Machine learning approaches that require data to be copied to a single location are hampered by the challenges of data sharing. Federated Learning (FL) is a promising approach to learn a joint model over all the available data across silos. In many cases, the sites participating in a federation have different data distributions and computational capabilities. In these heterogeneous environments existing approaches exhibit poor performance: synchronous FL protocols are communication efficient, but have slow learning convergence and high energy cost; conversely, asynchronous FL protocols have faster convergence with lower energy cost, but higher communication. In this work, we present a novel energy-efficient Semi-Synchronous Federated Learning protocol that mixes local models periodically with minimal idle time and fast convergence. We show through extensive experiments over established benchmark datasets in the computer-vision domain as well as in real-world biomedical settings that our approach significantly outperforms previous work in data and computationally heterogeneous environments.


 Jose Luis Ambite, Ph.D., is an Associate Research Professor of Computer Science and a Research Team Leader at the Information Sciences Institute, both at the University of Southern California. He is an expert on information integration, including query rewriting under constraints,  learning  schema  mappings,  and  entity  linkage.  His  current  research  is  on  biomedical  data science, developing novel approaches for integration and analysis of biomedical and genetic data within  several  large  NIH-funded  projects  (prisms-study.org,  nimhgenetics.org,  schizconnect.org, pagestudy.org, and bigdatau.org), including efficient federated learning for biomedical domains.


 Dimitris Stripelis, M.S., is a PhD candidate in Computer Science at the University of Southern California. Dimitris holds a BSc in Computer Science from the Athens University of Economics and Business and an MSc in Computer Science (Data Science specialization) from the University of Southern California. His research focus is on federated learning, federated data management systems, and data integration.


Host: Deborah Khider POC: Alma Nava

The Speaker approved to be recorded. The recording for this AI Seminar talk will be posted on our USC/ISI YouTube page within 1-2 business days: https://www.youtube.com/user/USCISI.

Subscribe here to learn more about upcoming seminars: https://dev.isi.edu/events/isi-seminar-series/


« Return to Upcoming Events