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:20260615T232728Z
UID:1781555248@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.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=7959&cHash=90
 cb5e34afc5d806c57c6b82702733c0\nAbst
 ract\n Machine learning is poised to
  play an increasingly central role i
 n the future of autonomous systems. 
 However, deploying learning-based me
 thods safely and reliably in the rea
 l world requires a principled and in
 tegrated theoretical understanding o
 f learning-based control. In this ta
 lk, I will present recent progress t
 oward this goal, drawing on tools fr
 om both systems theory and learning 
 theory. In the first part of the tal
 k, we will explore the fundamental l
 imits of learning-based control: wha
 t makes a system easy or hard to lea
 rn? Our focus will be on sample comp
 lexity - the minimum number of sampl
 es required to accurately learn a mo
 del or control policy. We will show 
 how system-theoretic properties such
  as controllability can significantl
 y influence the learning process. In
  particular, we will demonstrate tha
 t systems with poor controllability 
 structure - such as underactuated sy
 stems - can exhibit provably high sa
 mple complexity, regardless of the l
 earning algorithm used. Time permitt
 ing, the second part of the talk wil
 l discuss the use of online learning
  techniques for adaptive control in 
 dynamic environments. We will focus 
 on the problem of online tracking co
 ntrol of unknown and moving targets,
  which may be non-stationary and rev
 ealed only sequentially. By leveragi
 ng online learning methods, we can d
 esign control algorithms that come w
 ith theoretical performance guarante
 es despite the nonstationarity. We w
 ill demonstrate the practical effect
 iveness of these methods through exp
 eriments on a real quadrotor platfor
 m.\n \n About the Speaker\n Anastasi
 os Tsiamis received the Diploma degr
 ee in electrical and computer engine
 ering from the National Technical Un
 iversity of Athens, Greece, in 2014.
  He obtained his PhD at the Departme
 nt of Electrical and Systems Enginee
 ring, University of Pennsylvania, Ph
 iladelphia, PA, USA, in 2022. Curren
 tly, he is a senior scientist at the
  Automatic Control Laboratory, ETH Z
 urich, Switzerland. His research int
 erests include statistical and onlin
 e learning in the setting of control
  systems, as well as robust and risk
 -aware control. Anastasios Tsiamis w
 as a finalist for the IFAC Young Aut
 hor Prize in IFAC 2017 World Congres
 s and a finalist for the Best Studen
 t Paper Award in ACC 2019. He is a c
 oauthor to the paper that has won th
 e Best Student Paper Award in CDC 20
 22.\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