BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Sequential Decision Making in Experimental Design and Sustainabili
 ty via Adaptive Submodularity - Andreas Krause\, ETH Zurich
DTSTART:20120229T140000Z
DTEND:20120229T150000Z
UID:TALK36587@talks.cam.ac.uk
CONTACT:Microsoft Research Cambridge Talks Admins
DESCRIPTION:Solving sequential decision problems under partial observabili
 ty is a fundamental but notoriously difficult challenge. I will introduce 
 the new concept of adaptive submodularity\, generalizing the classical not
 ion of submodular set functions to adaptive policies. We prove that\, if a
  problem satisfies this property\, a simple adaptive greedy algorithm is g
 uaranteed to be competitive with the optimal policy. The concept allows us
  to recover\, generalize\, and extend existing results in diverse applicat
 ions\, including sensor management\, viral marketing\, and active learning
 . I will focus on two case studies. In an application to Bayesian experime
 ntal design in Behavioral Economics\, we show how greedy optimization of a
  novel adaptive submodular criterion outperforms standard myopic technique
 s based on information gain and value of information. I will also discuss 
 how adaptive submodularity can help to address problems in computational s
 ustainability by presenting results on conservation planning for three rar
 e species in the Pacific Northwest of the United States.\n\nThis talk is b
 ased on joint work primarily with Daniel Golovin
LOCATION:Small public lecture room\, Microsoft Research Ltd\, 7 J J Thomso
 n Avenue (Off Madingley Road)\, Cambridge
END:VEVENT
END:VCALENDAR