UT Austin Villa Publications

Sorted by DateClassified by Publication TypeClassified by TopicSorted by First Author Last Name

Deep Reinforcement Learning in Parameterized Action Space

Matthew Hausknecht and Peter Stone. Deep Reinforcement Learning in Parameterized Action Space. In Proceedings of the International Conference on Learning Representations (ICLR), May 2016.

Download

[PDF]468.3kB  

Abstract

Recent work has shown that deep neural networks are capable ofapproximating both value functions and policies in reinforcementlearning domains featuring continuous state and actionspaces. However, to the best of our knowledge no previous work hassucceeded at using deep neural networks in structured (parameterized)continuous action spaces. To fill this gap, this paper focuses onlearning within the domain of simulated RoboCup soccer, which featuresa small set of discrete action types, each of which is parameterizedwith continuous variables. The best learned agents can score goalsmore reliably than the 2012 RoboCup champion agent. As such, thispaper represents a successful extension of deep reinforcement learningto the class of parameterized action space MDPs.

BibTeX

@InProceedings{ICLR16-hausknecht,
  author = {Matthew Hausknecht and Peter Stone},
  title = {Deep Reinforcement Learning in Parameterized Action Space},
  booktitle = {Proceedings of the International Conference on Learning Representations (ICLR)},
  location = {San Juan, Puerto Rico},
  month = {May},
  year = {2016},
  abstract = {
Recent work has shown that deep neural networks are capable of
approximating both value functions and policies in reinforcement
learning domains featuring continuous state and action
spaces. However, to the best of our knowledge no previous work has
succeeded at using deep neural networks in structured (parameterized)
continuous action spaces. To fill this gap, this paper focuses on
learning within the domain of simulated RoboCup soccer, which features
a small set of discrete action types, each of which is parameterized
with continuous variables. The best learned agents can score goals
more reliably than the 2012 RoboCup champion agent. As such, this
paper represents a successful extension of deep reinforcement learning
to the class of parameterized action space MDPs.
  },
}

Generated by bib2html.pl (written by Patrick Riley ) on Thu Oct 12, 2017 13:11:26