BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Enabling Fast\, Robust\, and Personalized Federated Learning
DTSTART:20231129T161500
DTEND:20231129T171500
DTSTAMP:20260414T215410Z
UID:e484a878ed0af66bf7b3c01a3aa3b7d3eff9fe96c0f0a1c67aa2e818
CATEGORIES:Conferences - Seminars
DESCRIPTION:Prof. Ramtin Pedarsani Associate Professor\n​Electrical and
  Computer Engineering / UCSB\nIn many large-scale machine learning applica
 tions\, data is acquired and processed at the edge nodes of the network su
 ch as mobile devices\, users’ devices\, and IoT sensors. While distribut
 ed learning at the edge can enable a variety of new applications\, it face
 s major systems bottlenecks that severely limit its reliability and scalab
 ility including system and data heterogeneity and communication bottleneck
 . In this talk\, we focus on federated learning which is a new distributed
  machine learning approach\, where a model is trained over a set of device
 s such as mobile phones\, while keeping data localized. We first present a
  straggler-resilient federated learning scheme that uses adaptive node par
 ticipation to tackle the challenge of system heterogeneity. We next presen
 t a robust optimization formulation for federated learning that enables us
  to address the data heterogeneity challenge in federated learning. We fin
 ally talk about a new algorithm for personalizing the learned models for d
 ifferent users. 
LOCATION:BC 129 https://plan.epfl.ch/?room==BC%20129
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR