Predictive State Representations
The ambition of Predictive State Representations (PSRs) is to bypass the need for defining state spaces in reinforcement learning problems. The idea is that predictions of observation sequences conditioned on action sequences can serve as sufficient statistics for choosing (or learning to choose) optimal actions.
Peter Stone Faculty pstone [at] cs utexas edu
Towards Employing PSRs in a Continuous Domain 2004
Nicholas K. Jong and Peter Stone, Technical Report UT-AI-TR-04-309, The University of Texas at Austin, Department of Computer Sciences, AI Laboratory.
Learning Predictive State Representations 2003
Satinder Singh, Michael L. Littman, Nicholas K. Jong, David Pardoe, and Peter Stone, In Proceedings of the Twentieth International Conference on Machine Learning, August 2003.