Home

About

Advanced Search

Browse by Discipline

Scientific Societies

E-print Alerts

Add E-prints

E-print Network
FAQHELPSITE MAPCONTACT US


  Advanced Search  

 
CLUSTERING FINITE DISCRETE MARKOV CHAINS Greg Ridgeway, University of Washington and Steven Altschuler, Microsoft Corp.
 

Summary: CLUSTERING FINITE DISCRETE MARKOV CHAINS
Greg Ridgeway, University of Washington and Steven Altschuler, Microsoft Corp.
Greg Ridgeway, Box 354322, University of Washington, Seattle, WA 98195-4322
Keywords: clustering, Markov chain, mixture modeling
Abstract
In problem situations where observations consist of a
sequence of events, Markov models often prove useful.
However, when there is suspected heterogeneity among
the Markov transition kernels generating the observed
sequences, more refined methods become necessary. In
this paper we describe a probabilistic method for
clustering Markov processes with a pre-specified
number of clusters. We derive a Gibbs sampler and a
computationally efficient hybrid MCMC-constrained
EM algorithm.
Introduction
Consider an s-state discrete Markov process (Ross
[1993]) where the transition matrix for the process is
unknown. Further assume that a dataset of N such
processes, possibly of different length, exists in which

  

Source: Altschuler, Steve - Department of Pharmacology, UT Southwestern Medical Center

 

Collections: Biotechnology