Higher-Order Kullback-Leibler Aggregation of Markov Chains

Abstract

In this work, we consider the problem of reducing a first-order Markov chain on a large alphabet to a higher-order Markov chain on a small alphabet. We present informationtheoretic cost functions that are related to predictability and lumpability, show relations between these cost functions, and discuss heuristics to minimize them. Our experiments suggest that the generalization to higher orders is useful for model reduction in reliability analysis and natural language processing.

Extracted Key Phrases

4 Figures and Tables

Cite this paper

@article{Geiger2016HigherOrderKA, title={Higher-Order Kullback-Leibler Aggregation of Markov Chains}, author={Bernhard C. Geiger and Yuchen Wu}, journal={CoRR}, year={2016}, volume={abs/1608.04637} }