BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Enabling Fast\, Robust\, and Personalized Federated Learning - Pro
 f Ramtin Pedarsani\, UC Santa Barbara 
DTSTART:20240207T140000Z
DTEND:20240207T150000Z
UID:TALK208732@talks.cam.ac.uk
CONTACT:Dr Varun Jog
DESCRIPTION:In many large-scale machine learning applications\, data is ac
 quired and processed at the edge nodes of the network such as mobile devic
 es\, users’ devices\, and IoT sensors. While distributed learning at the
  edge can enable a variety of new applications\, it faces major systems bo
 ttlenecks that severely limit its reliability and scalability including sy
 stem and data heterogeneity and communication bottleneck. In this talk\, w
 e focus on federated learning which is a new distributed machine learning 
 approach\, where a model is trained over a set of devices such as mobile p
 hones\, while keeping data localized. We first present a straggler-resilie
 nt federated learning scheme that uses adaptive node participation to tack
 le the challenge of system heterogeneity. We next present a robust optimiz
 ation formulation for federated learning that enables us to address the da
 ta heterogeneity challenge in federated learning. We finally talk about a 
 new algorithm for personalizing the learned models for different users.
LOCATION:MR5\, CMS Pavilion A
END:VEVENT
END:VCALENDAR