BEGIN:VCALENDAR
PRODID:-//eluceo/ical//2.0/EN
VERSION:2.0
CALSCALE:GREGORIAN
BEGIN:VEVENT
UID:95021017b02e89d4838e6d683fe41c36
DTSTAMP:20260421T194402Z
SUMMARY:LC2 Seminar –  Cesare Molinari (Università di Genova)
DESCRIPTION:Title: Learning from data via overparameterization\nAbstract: S
 olving data-driven problems requires defining complex models and fitting t
 hem on data\, neural networks being a motivating example. The fitting proc
 edure can be seen as an optimization problem\, which is often non-convex\,
  and hence optimization guarantees are hard to derive. An opportunity is p
 rovided by viewing the model of interest as a redundant reparameterization
 —an overparameterization—of some simpler model for which optimization 
 results are easier to achieve. In this talk\, after formalizing the above 
 idea\, we review some recent results and derive new ones. In particular\, 
 we consider the gradient flow of some classes of linear overparameterizati
 on and show they correspond to suitable mirror flow on the original parame
 ters. Our main contribution relates to the study of the latter\, for which
  we establish well-posedness and convergence. The results yield insight on
  the role of overparameterization for implicit regularization and constrai
 ned optimization.
URL:https://www.imperial.ac.uk/events/205094/lc2-seminar-cesare-molinari-un
 iversita-di-genova/
DTSTART;TZID=Europe/London:20260304T120000
DTEND;TZID=Europe/London:20260304T130000
LOCATION:145\, Huxley Building\, South Kensington Campus\, Imperial College
  London\, London\, SW7 2AZ\, United Kingdom
END:VEVENT
BEGIN:VTIMEZONE
TZID:Europe/London
BEGIN:STANDARD
DTSTART:20260304T120000
TZNAME:GMT
TZOFFSETTO:+0000
TZOFFSETFROM:+0000
END:STANDARD
END:VTIMEZONE
END:VCALENDAR