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:20250717T111231Z
LAST-MODIFIED:20250717T111231Z
DTSTAMP:20260719T220618Z
UID:1784487978@tuc.gr
SUMMARY:"Learning Theory for Control: Algori
 thms, Rates, Fundamental Limits" by 
 Dr. Anastasios Tsiamis
LOCATION:Λ - Κτίριο Επιστημών/ΗΜΜΥ, 145Π-42
DESCRIPTION:https://www.ece.tuc.gr/el/katalogos-
 ekdiloseon?tx_tucevents2_tuceventsdi
 splay%5Baction%5D=show&tx_tucevents2
 _tuceventsdisplay%5Bcontroller%5D=Ev
 ent&tx_tucevents2_tuceventsdisplay%5
 Bevent%5D=7959&cHash=bd4fb59cba07a3a
 aeb2d50f52488a32d\nAbstract\n Machin
 e learning is poised to play an incr
 easingly central role in the future 
 of autonomous systems. However, depl
 oying learning-based methods safely 
 and reliably in the real world requi
 res a principled and integrated theo
 retical understanding of learning-ba
 sed control. In this talk, I will pr
 esent recent progress toward this go
 al, drawing on tools from both syste
 ms theory and learning theory. In th
 e first part of the talk, we will ex
 plore the fundamental limits of lear
 ning-based control: what makes a sys
 tem easy or hard to learn? Our focus
  will be on sample complexity - the 
 minimum number of samples required t
 o accurately learn a model or contro
 l policy. We will show how system-th
 eoretic properties such as controlla
 bility can significantly influence t
 he learning process. In particular, 
 we will demonstrate that systems wit
 h poor controllability structure - s
 uch as underactuated systems - can e
 xhibit provably high sample complexi
 ty, regardless of the learning algor
 ithm used. Time permitting, the seco
 nd part of the talk will discuss the
  use of online learning techniques f
 or adaptive control in dynamic envir
 onments. We will focus on the proble
 m of online tracking control of unkn
 own and moving targets, which may be
  non-stationary and revealed only se
 quentially. By leveraging online lea
 rning methods, we can design control
  algorithms that come with theoretic
 al performance guarantees despite th
 e nonstationarity. We will demonstra
 te the practical effectiveness of th
 ese methods through experiments on a
  real quadrotor platform.\n \n About
  the Speaker\n Anastasios Tsiamis re
 ceived the Diploma degree in electri
 cal and computer engineering from th
 e National Technical University of A
 thens, Greece, in 2014. He obtained 
 his PhD at the Department of Electri
 cal and Systems Engineering, Univers
 ity of Pennsylvania, Philadelphia, P
 A, USA, in 2022. Currently, he is a 
 senior scientist at the Automatic Co
 ntrol Laboratory, ETH Zurich, Switze
 rland. His research interests includ
 e statistical and online learning in
  the setting of control systems, as 
 well as robust and risk-aware contro
 l. Anastasios Tsiamis was a finalist
  for the IFAC Young Author Prize in 
 IFAC 2017 World Congress and a final
 ist for the Best Student Paper Award
  in ACC 2019. He is a coauthor to th
 e paper that has won the Best Studen
 t Paper Award in CDC 2022.\n
STATUS:CONFIRMED
ORGANIZER;RSVP=FALSE;CN=TUC;CUTYPE=TUC:mailto:webmaster@tuc.gr
DTSTART:20250721T110000
DTEND:20250721T120000
TRANSP:OPAQUE
CLASS:DEFAULT
END:VEVENT
END:VCALENDAR