BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Minimax-Bayes Reinforcement Learning
DTSTART:20230901T110000
DTEND:20230901T120000
DTSTAMP:20260505T131129Z
UID:e4f3f1e4d0d553c1cd57a522f412adc0dbe16b2e2e96ffd20dad1479
CATEGORIES:Conferences - Seminars
DESCRIPTION:Prof. Christos Dimitrakakis\, University of St. Gallen\, Switz
 erland\nAbstract:\nWhile the Bayesian decision-theoretic framework offers 
 an elegant solution to the problem of decision making under uncertainty\,
  one question is how to appropriately select the prior distribution. One
  idea is to employ a worst-case prior. However\, this is not as easy to 
 specify in sequential decision making as in simple statistical estimation
  problems.  I will present (sometimes approximate) minimax-Bayes solutio
 ns for various reinforcement learning problems to gain insights into the 
 properties of the corresponding priors and policies. We find that while t
 he worst-case prior depends on the setting\, the corresponding minimax po
 licies are more robust than those that assume a standard (i.e.\\ uniform)
  prior.\n\nBio:\nMy research in artificial intelligence and machine learni
 ng includes reinforcement learning\, preference elicitation\, privacy\, fa
 irness\, experiment design\, safety\, recommendation systems\, job matchin
 g\, crowdsourcing and social computing more generally. More recently\, I h
 ave been interested in the computational\, societal and statistical proble
 ms that arise when humans interact with artificially intelligent systems. 
 I obtained my PhD in 2006 from EPFL\, and I am currently a professor a the
  university of St. Gallen. Previously\, I was a professor at the universit
 y of Oslo\, an assistant professor at the university of Lille and a senior
  researcher at Chalmers university.\n 
LOCATION:ME C2 405
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR