Matthew E. Taylor's Publications

Sorted by DateClassified by Publication TypeSorted by First Author Last NameClassified by Research Category

Integrating Human Demonstration and Reinforcement Learning: Initial Results in Human-Agent Transfer

Matthew E. Taylor and Sonia Chernova. Integrating Human Demonstration and Reinforcement Learning: Initial Results in Human-Agent Transfer. In Proceedings of the Agents Learning Interactively from Human Teachers workshop (at AAMAS-10), May 2010.
ALIHT-10

Download

[PDF]142.6kB  [postscript]235.7kB  

Abstract

This work introduces Human-Agent Transfer (HAT), a method thatcombines transfer learning, learning from demonstration andreinforcement learning to achieve rapid learning and high performancein complex domains. Using experiments in a simulated robot soccerdomain, we show that human demonstrations can be transferred into abaseline policy for an agent, and reinforcement learning can be usedto significantly improve policy performance. These results are animportant initial step that suggest that agents can not only quicklylearn to mimic human actions, but that they can also learn to surpassthe abilities of the teacher.

BibTeX Entry

@inproceedings(ALIHT10-Taylor,
  author="Matthew E.\ Taylor and Sonia Chernova",
  title="Integrating Human Demonstration and Reinforcement Learning: Initial Results in Human-Agent Transfer"
  Booktitle="Proceedings of the Agents Learning Interactively from Human Teachers workshop (at AAMAS-10)",
  month="May",
  year= "2010",
  wwwnote={<a href="www.cs.utexas.edu/~bradknox/AAMAS-ALIHT/">ALIHT-10</a>}
  abstract={
This work introduces Human-Agent Transfer (HAT), a method that
combines transfer learning, learning from demonstration and
reinforcement learning to achieve rapid learning and high performance
in complex domains. Using experiments in a simulated robot soccer
domain, we show that human demonstrations can be transferred into a
baseline policy for an agent, and reinforcement learning can be used
to significantly improve policy performance.  These results are an
important initial step that suggest that agents can not only quickly
learn to mimic human actions, but that they can also learn to surpass
the abilities of the teacher.}
)

Generated by bib2html.pl (written by Patrick Riley ) on Mon Apr 19, 2010 14:12:46