BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Memento EPFL//
BEGIN:VEVENT
SUMMARY:Finding a most biased coin with fewest flips
DTSTART:20140725T150000
DTEND:20140725T160000
DTSTAMP:20260509T170111Z
UID:16f81608749cac2c77c715e7e84682cae0cf1913cae47090bb77437c
CATEGORIES:Conferences - Seminars
DESCRIPTION:Karthik Chandrasekaran\, Harvard\nAbstract: The multi-armed b
 andit problem is a well-studied problem with applications in bioinformati
 cs\, clinical trials\, etc. A variation of the problem is to find the mos
 t rewarding arm in the fewest possible number of steps. When the rewards 
 are Bernoulli\, this is equivalent to the problem of finding the most bia
 sed coin among a set of coins by tossing them adaptively. The goal here i
 s to devise a strategy that minimizes the expected number of tosses until
  there exists a coin whose posterior probability of being most biased is 
 at least 1-delta\, for a given delta. In this talk\, I will present an op
 timal adaptive strategy for a particular probabilistic setting of the pro
 blem. I will\nshow that the strategy performs the best possible action in 
 each step by employing tools from the field of Markov games.\nBased on jo
 int work with Richard Karp.\nBio: Karthik Chandrasekaran is a Simons post
 doctoral research fellow at Harvard University. He obtained his B. Tech. 
 in Computer Science and Engineering from the Indian Institute of Technolo
 gy\, Madras and his Ph.D. in Algorithms\, Combinatorics\, and Optimizatio
 n from Georgia Tech. His primary research interests are in optimization\,
  integer programming\, probabilistic methods and analysis\, and randomize
 d algorithms. He will be joining as an assistant professor in the Univer
 sity of Illinois at Urbana-Champaign in Oct\, 2014.
LOCATION:BC 420 https://plan.epfl.ch/?room==BC%20420
STATUS:CONFIRMED
END:VEVENT
END:VCALENDAR