The Imitation Page
For some time, I have been focused on developing formal models of imitation
which allow an agent to observe others, extract useful information
about behaviours from these agents and integrate this information
into their own behavioural repertoires. Key problems we have dealt
with in this domain include assessing the relative quality
of information observed from others versus the information already
at hand, dealing with heterogeneous action capabilities and imitation in
complex domains requiring real-valued attributes, large state
spaces and generalization. Imitation is of course closely related to
Learning Apprentices, Programming by Demonstration, Learning by Demonstration,
Learning from Demonstration and Behavioural Cloning (or Behavioral Cloning).
Imitation Learning Papers
Bob Price and Craig Boutilier, Implicit Imitation in Multi-agent Reinforcement
Learning, ICML '99, International Conference on Machine Learning.
(200k
g-zipped Postscript)
Bob Price and Craig Boutilier, Imitation and Reinforcement learning
in agents with herterogeneous actions. AISB '00, (90k
g-ziped Postscript)
Imitation Around the World
This section concerned primarily with computational models of imitation
or theoretical frameworks for imitation that are useful in thinking about
computation. Some material related to natural forms of imitation appear
where it introduces interesting questions about the nature of imitation,
but this site is not intended to be an authoritative index of natural forms
of imitation.
October
18, 2001