BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Data-driven Analysis of First-Order Methods via Distributionally R
 obust Optimization
DTSTART:20260106T110000
DTEND:20260106T120000
DTSTAMP:20260407T111215Z
UID:8a98bb0de2552a71d6eaa43e6cfa10807fdc7aa4e66dabed9deca55e
CATEGORIES:Conferences - Seminars
DESCRIPTION:Bartolomeo Stellato Assistant Professor Princeton University O
 perations Research and Financial Engineering\nAbstract\nWe consider the pr
 oblem of analyzing the probabilistic performance of first-order methods wh
 en solving convex optimization problems drawn from an unknown distribution
  only accessible through samples. By combining performance estimation (PEP
 ) and Wasserstein distributionally robust optimization (DRO)\, we formulat
 e the analysis as a tractable semidefinite program. Our approach unifies w
 orst-case and average-case analyses by incorporating data-driven informati
 on from the observed convergence of first-order methods on a limited numbe
 r of problem instances. This yields probabilistic\, data-driven performanc
 e guarantees in terms of the expectation or conditional value-at-risk of t
 he selected performance metric. Experiments on smooth convex minimization\
 , logistics regression\, and Lasso show that our method significantly redu
 ces the conservatism of classical worst-case bounds and narrows the gap be
 tween theoretical and empirical performance.\n\n\nBiography \nAssistant P
 rofessor at the Department of Operations Research and Financial Engineeri
 ng at Princeton University. He is associated faculty in the Department 
 of Electrical and Computer Engineering and the Department of Computer Sc
 ience\, and an affiliated faculty in the AI at Princeton initiative\, th
 e Princeton Institute for Computational Science and Engineering\, the Ce
 nter for Statistics and Machine Learning\, and the Robotics at Princeton
  initiative. Prof Stellato is also a fellow at the Princeton Whitman an
 d Yeh colleges.\nHis research lies at the interface of mathematical op
 timization\, machine learning\, and optimal control. It focuses on data-dr
 iven computational tools to make decisions in highly dynamic and uncertain
  environments.\n\n 
LOCATION:ME C2 405 https://plan.epfl.ch/?room==ME%20C2%20405
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR