Search Machine Learning Repository:
Modelling Sparse Dynamical Systems with Compressed Predictive State Representations
Authors: William L. Hamilton, Mahdi M. Fard and Joelle Pineau
Conference: Proceedings of the 30th International Conference on Machine Learning (ICML-13)
Abstract: Efficiently learning accurate models of dynamical systems is of central importance for developing rational agents that can succeed in a wide range of challenging domains. The difficulty of this learning problem is particularly acute in settings with large observation spaces and partial observability. We present a new algorithm, called Compressed Predictive State Representation (CPSR), for learning models of high-dimensional partially observable uncontrolled dynamical systems from small sample sets. The algorithm, which extends previous work on Predictive State Representations, exploits a particular sparse structure present in many domains. This sparse structure is used to compress information during learning, allowing for an increase in both the efficiency and predictive power. The compression technique also relieves the burden of domain specific feature selection and allows for domains with extremely large discrete observation spaces to be efficiently modelled. We present empirical results showing that the algorithm is able to build accurate models more efficiently than its uncompressed counterparts, and provide theoretical results on the accuracy of the learned compressed model.
authors venues years
Suggest Changes to this paper.
Brought to you by the WUSTL Machine Learning Group. We have open faculty positions (tenured and tenure-track).