UT ML Group: Reinforcement Learning

Reinforcement Learning (RL) consists of a set of machine learning methods that address a particular kind of learning task, in which the learner is placed in an unknown environment and is allowed to take actions that bring it rewards and can change its state in the environment. In general terms, the goal of the agent is to develop a policy, or a mapping from states to actions, that maximizes the reward it obtains while interacting with the environment.

Existing RL methods require no human involvement but may take a large amount of interaction with the environment to develop an effective policy. Our research has focused on aiding the learner by providing advice and allowing it to select the parts of the environment in which additional experience is needed.

Publications

  1. Using Active Relocation to Aid Reinforcement Learning [Abstract] [PDF]
    Lilyana Mihalkova and Raymond Mooney
    In Proceedings of the 19th International FLAIRS Conference (FLAIRS-2006), pp. 580-585, Melbourne Beach, Florida, May 2006.

  2. Guiding a Reinforcement Learner with Natural Language Advice: Initial Results in RoboCup Soccer [Abstract] [PDF]
    Gregory Kuhlmann, Peter Stone, Raymond J. Mooney, and Jude W. Shavlik
    Proceedings of the AAAI-2004 Workshop on Supervisory Control of Learning and Adaptive Systems, pp. 30-35, San Jose, CA, July 2004.

mooney@cs.utexas.edu