BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//TUC//Events//EN
CALSCALE:GREGORIAN
BEGIN:VTIMEZONE
TZID:Europe/Athens
TZNAME:EEST
DTSTART:19700329T030000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=3
BEGIN:STANDARD
TZOFFSETFROM:+0200
TZOFFSETTO:+0300
TZNAME:EET
DTSTART:19701025T040000
RRULE:FREQ=YEARLY;BYDAY=-1SU;BYMONTH=10
END:STANDARD
END:VTIMEZONE
BEGIN:VEVENT
CREATED:20231027T095603Z
LAST-MODIFIED:20231027T095603Z
DTSTAMP:20260809T194051Z
UID:1786293651@tuc.gr
SUMMARY:Ομιλία κου Βασίλη Απιδόπουλου "Itera
 tive regularization for classificati
 on via hinge loss dual diagonal desc
 ent"
LOCATION:
DESCRIPTION:https://www.tuc.gr/el/to-polytechnei
 o/ilektronikes-ypiresies/imerologio/
 imerologio-ekdiloseon-1?tx_tucevents
 2_tuceventsdisplay%5Baction%5D=show&
 tx_tucevents2_tuceventsdisplay%5Bcon
 troller%5D=Event&tx_tucevents2_tucev
 entsdisplay%5Bevent%5D=6585&cHash=4e
 04f7579019263849f9235377476484\nAbst
 ract\n Iterative (implicit) regulari
 zation is a classic idea in regulari
 zation theory that has recently beco
 me popular in machine learning. On t
 he one hand, it allows to design eff
 icient algorithms controlling at the
  same time numerical and statistical
  accuracy. On the other hand, it all
 ows shedding light on the learning c
 urves observed while training neural
  networks. In this talk, we will foc
 us on iterative regularization in th
 e context of classification. After c
 ontrasting this setting with regress
 ion and inverse problems, we develop
  an iterative regularization approac
 h based on the hinge loss function, 
 used frequently in practice. More pr
 ecisely we consider a diagonal appro
 ach for a family of algorithms for w
 hich we prove convergence as well as
  rates of convergence. Our approach 
 compares favorably with other altern
 atives, as confirmed also in numeric
 al simulations.\n \n About the Speak
 er\n Vassilis Apidopoulos is a Post-
 Doctoral Researcher in the Laborator
 y for Computational and Statistical 
 Learning (LCSL) at MaLGa Research Ce
 nter (Università di Genova), working
  with Silvia Villa and Lorenzo Rosas
 co on implicit regularization. He co
 mpleted his Ph.D. at the Institut de
  Mathématiques de Bordeaux, based on
  the study of inertial gradient desc
 ent algorithms. Before that, he did 
 his Master’s Studies at the Universi
 té Claude Bernard Lyon 1 and the Éco
 le Normale Supérieure de Lyon, and h
 is undergraduate studies were comple
 ted in the Department of Mathematics
  of Aristotle University of Thessalo
 niki in Greece. His research interes
 ts lie in the field of optimization 
 with a particular focus on Machine l
 earning applications. \n
STATUS:CONFIRMED
ORGANIZER;RSVP=FALSE;CN=TUC;CUTYPE=TUC:mailto:webmaster@tuc.gr
DTSTART:20231110T173000
DTEND:20231110T183000
TRANSP:OPAQUE
CLASS:DEFAULT
END:VEVENT
END:VCALENDAR