Master's Defense

Active Exploration for Learning a Symbolic Representation

Speaker:Garrett Andersen
garrett at cs.duke.edu
Date: Friday, November 18, 2016
Time: 1:00pm - 3:00pm
Location: D344 LSRC, Duke

Abstract

We introduce an active exploration algorithm for efficiently learning an abstract symbolic model of an environment. We show that it outperforms random and greedy exploration policies on an Asteroids-inspired computer game domain.
Advisor(s): Ron Parr and George Konidaris
Committee: Carlo Tomasi, Kris Hauser