BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Learning Under Constraints: From Federated Collaboration to Black-
 Box LLMs - Salma Kharrat\, Kaust
DTSTART:20250804T100000Z
DTEND:20250804T110000Z
UID:TALK233434@talks.cam.ac.uk
CONTACT:Sally Matthews
DESCRIPTION:In both federated learning (FL) and large language model (LLMs
 ) optimization\, a central challenge is effective learning under constrain
 ts\, ranging from data heterogeneity and personalization to limited commun
 ication and black-box access. In this talk\, I present three approaches th
 at address these challenges across different settings. FilFL improves gene
 ralization in FL by filtering clients based on their joint contribution to
  global performance. DPFL tackles decentralized personalization by learnin
 g asymmetric collaboration graphs under strict resource budgets. Moving be
 yond FL\, I will present ACING\, a reinforcement learning method for optim
 izing instructions in black-box LLMs under strict query budgets\, where we
 ights and gradients are inaccessible. While these works tackle distinct pr
 oblems\, they are unified by a common goal: developing efficient learning 
 mechanisms that perform reliably under real-world constraints.
LOCATION:Computer Lab\, FW26
END:VEVENT
END:VCALENDAR