BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Learning to Optimize with Confidence
DTSTART:20160923T101500
DTEND:20160923T111500
DTSTAMP:20260407T081553Z
UID:93c0a13056696ffc2695190586ceb7a676ee52d4e3354355d0984281
CATEGORIES:Conferences - Seminars
DESCRIPTION:Prof. Andreas Krause\, ETHZ\nBio: Andreas Krause is an Associa
 te Professor of Computer Science at ETH Zurich\, where he leads the Learni
 ng & Adaptive Systems Group. Before that he was an Assistant Professor of 
 Computer Science at Caltech. He received his Ph.D. in Computer Science fro
 m Carnegie Mellon University and his Diplom in Computer Science and Mathem
 atics from TU Munich\, Germany. He is a Microsoft Research Faculty Fellow 
 and received an ERC Starting Investigator grant\, an NSF CAREER award as w
 ell as several best paper awards at premier conferences and journals.\nWit
 h the success of machine learning\, we increasingly see learning algorithm
 s make decisions in the real world. Often\, however\, this is in stark con
 trast to the classical train-test paradigm\, since the learning algorithm 
 affects the very data it must operate on.  A central challenge is to trad
 e exploration — collecting data for the sake of fitting better models 
 — and exploitation — using the estimate to make decisions.  In many a
 pplications\, such as in robotics\, exploration is a potentially dangerous
  proposition\, as it requires experimenting with actions with unknown cons
 equences.  In this talk\, I will formalize the problem of safe exploratio
 n as one of optimizing an unknown function subject to unknown constraints.
   Both objective and constraints are revealed through noisy experiments\,
  and safety requires that no infeasible action is chosen at any point. I w
 ill present an approach that uses Bayesian inference over the objective an
 d constraints\, which -- under some regularity assumptions -- is guarantee
 d to converge to a natural notion of reachable optimum. I will also show e
 xperiments on safe automatic parameter tuning of a robotic platform.\n 
LOCATION:ME C2 405 http://plan.epfl.ch/?zoom=20&recenter_y=5864084.17342&r
 ecenter_x=730960.62257&layerNodes=fonds\,batiments\,labels\,information\,p
 arkings_publics\,arrets_metro\,transports_publics&floor=2&q=me_c2%20405
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR