BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Talks.cam//talks.cam.ac.uk//
X-WR-CALNAME:Talks.cam
BEGIN:VEVENT
SUMMARY:Rational learning may be minimalist - Paul R. Schrater
DTSTART:20120424T140000Z
DTEND:20120424T150000Z
UID:TALK37824@talks.cam.ac.uk
CONTACT:Dr. Cristina Savin
DESCRIPTION:What sorts of things should you learn about the environment?  
 To be\nconcrete\, if I expose you to the projectile motions in a game like
 \nAngry Birds\, how much will you learn about the trajectories of the\nbir
 ds? At one extreme\, you might use extensive feedback to hone\nstrategies 
 for controlling the bird’s destructive desires without\nunderstanding th
 e details of the trajectory\, and at the other extreme\,\nyou may acquire 
 a highly accurate predictive model for trajectories\nthat can allow for co
 mplex and novel interactions\, like mid-flight\ncontrol.  Current learning
  theory offers little guidance to predict\nwhat aspects of the environment
  will be learned\, and what sorts of\ntask and feedback will facilitate or
  inhibit learning richer internal\nmodels.  I will describe experimental r
 esults from our lab that\nsupport a kind of minimalist learning strategy\,
  we could phrase “only\nlearn what you need”.  Minimalist learning pre
 dicts that internal\nmodels will only be acquired when both the task requi
 res\nprediction/counterfactual reasoning\, and that model improvement can 
 be\nanticipated to improve performance.   I show counter-intuitive\nexperi
 mental evidence for this hypothesis in a family of angry\nbird-like tasks\
 , including more learning with less-reliable data\, no\nlearning with full
  feedback\, and how subtle changes to the predictive\nrequirements in a ta
 sk can lead to large differences in internal model\nlearning.   I will sho
 w how modeling learning from a Bayesian adaptive\ncontrol perspective with
  cognitive costs can provide a normative\nframework for miminalist learnin
 g\, and will argue that miminalist\nlearning may be critical for skill for
 mation.
LOCATION:Cambridge University Engineering Department\, CBL Rm #438 (http:/
 /learning.eng.cam.ac.uk/Public/Directions)
END:VEVENT
END:VCALENDAR