 
Summary: COLLOQUIUM
University of Regina
Department of Mathematics and Statistics
Speaker: Sandra Zilles (University of Regina)
Title: Learning in groupstructured action/state spaces
Time & Place: Friday, September 25, 3:30  4:30 pm, CL 126
Abstract
The presentation addresses the problem of actively learning the transition
dynamics of deterministic, discretestate environments. We assume that
an agent exploring such an environment is able to perform actions in the
environment and to sense the state changes. The question investigated is
whether the agent can learn the dynamics without visiting all states. Such
a goal is unrealistic in general, hence we assume that the environment has
structural properties an agent might exploit. In particular, we assume that
the set of all action sequences forms an algebraic group.
We introduce a learning model in different variants and study under
which circumstances group structures can be learned efficiently from exper
imenting with group generators (actions). It turns out that for some classes
of such environments the choice of actions given to the agent determines if
efficient learning is possible. Negative results are presented, even without
