neural networks research group
areas
people
projects
demos
publications
software/data
Comparing Evolutionary and Temporal Difference Methods for Reinforcement Learning (2006)
Matthew Taylor
and
Shimon Whiteson
and
Peter Stone
Both genetic algorithms (GAs) and temporal difference (TD) methods have proven effective at solving reinforcement learning (RL) problems. However, since few rigorous empirical comparisons have been conducted, there are no general guidelines describing the methods' relative strengths and weaknesses. This paper presents the results of a detailed empirical comparison between a GA and a TD method in Keepaway, a standard RL benchmark domain based on robot soccer. In particular, we compare the performance of NEAT~citestanley:ec02evolving, a GA that evolves neural networks, with Sarsa~citeRummery94, Singh96, a popular TD method. The results demonstrate that NEAT can learn better policies in this task, though it requires more evaluations to do so. Additional experiments in two variations of Keepaway demonstrate that Sarsa learns better policies when the task is fully observable and NEAT learns faster when the task is deterministic. Together, these results help isolate the factors critical to the performance of each method and yield insights into their general strengths and weaknesses.
View:
PDF
,
PS
,
HTML
Citation:
In
Proceedings of the Genetic and Evolutionary Computation Conference
, 1321-28, July 2006.
Bibtex:
@InProceedings{GECCO06-matt, title={Comparing Evolutionary and Temporal Difference Methods for Reinforcement Learning}, author={Matthew Taylor and Shimon Whiteson and Peter Stone}, booktitle={Proceedings of the Genetic and Evolutionary Computation Conference}, month={July}, pages={1321-28}, url="http://nn.cs.utexas.edu/?GECCO06-matt", year={2006} }
People
Peter Stone
pstone [at] cs utexas edu
Matthew Taylor
taylorm [at] eecs wsu edu
Shimon Whiteson
Former Collaborator
s a whiteson [at] uva nl
Areas of Interest
Simulated Robot Soccer
Neuroevolution
Reinforcement Learning
Other Areas