 |
Comparing and Evaluating HMM Ensemble Training Algorithms Using Train and Test and Condition Number Criteria
Davis, Richard I. A. and Lovell, Brian C. (2004-02-01) Comparing and Evaluating HMM Ensemble Training Algorithms Using Train and Test and Condition Number Criteria. Pattern Analysis and Applications, 6 4: 327-335.
|
|
| |
| Attached Files (Some files may be inaccessible until you login with your UQ eSpace credentials) |
| Name |
Description |
MIMEType |
Size |
Downloads |
Comparing_and_evaluating_HMM.pdf
|
Comparing and evaluating HMM.pdf |
application/pdf |
256.53KB |
35 |
|
hmm_multicopy_fi.pdf
|
hmm_multicopy_fi.pdf |
application/pdf |
167.55KB |
919 |
| Author(s) |
Davis, Richard I. A.
Lovell, Brian C.
|
| Title |
Comparing and Evaluating HMM Ensemble Training Algorithms Using Train and Test and Condition Number Criteria
|
| Journal name |
Pattern Analysis and Applications
|
| Publication date |
2004-02-01
|
| Volume number |
6
|
| Issue number |
4
|
| ISSN |
1433-7541
|
| Start page |
327
|
| End page |
335
|
| Total pages |
10
|
| Editor(s) |
S. Singh
|
| Place of publication |
New York
|
| Publisher |
Springer
|
| Collection year |
2003
|
| Language |
eng
|
| Subject |
280207 Pattern Recognition
C1
280208 Computer Vision
700199 Computer software and services not elsewhere classified
|
| Abstract |
Hidden Markov Models have many applications in signal processing and pattern recognition, but their convergence-based training algorithms are known to suffer from over-sensitivity to the initial random model choice. This paper describes the boundary between regions in which ensemble learning is superior to Rabiner's multiplesequence Baum-Welch training method, and proposes techniques for determining the best method in any arbitrary situation. It also studies the suitability of the training methods using the condition number, a recently proposed diagnostic tool for testing the quality of the model. A new method for training Hidden Markov Models called the Viterbi Path counting algorithm is introduced and is found to produce significantly better performance than current methods in a range of trials.
|
| Keyword(s) |
iris-research
ensemble learning
Baum-Welch Algorithm
hidden Markov model
multiple sequence
model structure
Viterbi
|
| Additional Notes |
Originally published as Davis, Richard I. A. and Lovell, Brian C. (2004) Comparing and Evaluating HMM Ensemble Training Algorithms Using Train and Test and Condition Number Criteria. Pattern Analysis and Applications 6(4):327-336. doi: 10.1007/s10044-003-0198-6 Copyright 2004 Springer-Verlag. All rights reserved. The original article is available from www.springerlink.com PAA is rated as a Tier 1 journal at UQ. This paper describes fundamental work on pattern recognition that was started while I was on sabbatical at the University of Alberta with Terry Caelli. The collaboration with Terry Caelli resulted in several publications. Now Terry and I are both research leaders in NICTA and have continued our collaboration.
|
|
|
|
