|
Activity Number:
|
317
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Tuesday, August 5, 2008 : 2:00 PM to 3:50 PM
|
|
Sponsor:
|
Section on Statistical Computing
|
| Abstract - #301327 |
|
Title:
|
A Unified Approach to Computing Distributions Associated with Hidden State Sequences
|
|
Author(s):
|
Donald E.K. Martin*+ and John A.D. Aston
|
|
Companies:
|
North Carolina State University and Warwick University
|
|
Address:
|
209-E Patterson Hall, 2501 Founders Drive,, Raleigh, NC, 27695-8203,
|
|
Keywords:
|
Auxiliary Markov chain ; classification ; deterministic finite automaton ; distribution of pattern statistics ; hidden state sequences
|
|
Abstract:
|
In this paper a method is given for computing distributions of statistics of hidden state sequences. The method applies to any situation for which the conditional distribution of states given observations may be modeled by a factor graph with factors that depend on current and past states but not future ones. Model structure is exploited to develop a Markov chain that facilitates efficient computation of distributions. The methodology may be used for discrete hidden state sequences perturbed by noise and/or missing values, and for state sequences that serve to classify observations. A detailed example is given to illustrate the computational procedure.
|
