BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Game-theoretic logic learning in scientific domains - Stephen Mugg
 leton - Imperial College
DTSTART:20100727T130000Z
DTEND:20100727T140000Z
UID:TALK25549@talks.cam.ac.uk
CONTACT:Microsoft Research Cambridge Talks Admins
DESCRIPTION:*Abstract:* This talk is motivated by an attempt to model the 
 process of automated\nscientific discovery in terms of the theory of compe
 titive and collaborative\ngames. For instance\, at the object-level within
  Systems Biology interactions\nbetween a host and pathogen can be modelled
  as a form of adaptive competition.\nConversely at the meta-level groups o
 f experimental scientists can be viewed\nas conducting a collaborative gam
 e involving the proposal and refutation of\nhypotheses by experimentation.
  Some relevant concepts from Game Theory are\nbriefly reviewed. We then in
 troduce a formalism called Game-Theoretic Logic\nPrograms (GTLPs)\, which 
 allow modelling of multi-player strategies based on\nan adaptation of McCa
 rthy`s situation calculus. Lastly we propose an approach\nto machine learn
 ing such strategies. We argue that the approach\, called\nPre-emptive Stra
 tegy Learning (PSL)\, represents a departure from traditional\nforms of Ma
 chine Learning.  Usually Machine Learning is conceived in terms of\nextrac
 ting patterns from a database of past experiences with the aim of\npredict
 ing future outcomes. By contrast\, experimental choices in new areas\nof s
 cience might be compared to the problem BP faced during the oil spillage\n
 in the Caribbean.  The absence of relevent historical data dictates a new\
 napproach.  It is envisaged that PSL will machine learn strategies by samp
 ling\nprojected future events from a GTLP description of actions with asso
 ciated\nprobabilities and costs. In this way\, machine learning could be a
 pplied\niteratively within cycles of experiment planning. Although the dev
 elopment\nof GTLPs and PSL would have direct and immediate effect within a
 utomated\nscientific discovery tasks\, we believe that the approach should
  also have\nbroader application within other areas of computer science in 
 which\nsituations and actions have associated uncertainties and costs.
LOCATION:Small public lecture room\, Microsoft Research Ltd\, 7 J J Thomso
 n Avenue (Off Madingley Road)\, Cambridge
END:VEVENT
END:VCALENDAR