Support Vector Random Fields for Spatial Classification
Full Text:
2005_PKDD.pdf
In this paper we propose Support Vector Random Fields (SVRFs), an extension of Support Vector Machines (SVMs) that allows the modelling of spatial correlations in multi-dimensional data. SVRFs are derived as Conditional Random Fields that take advantage of the generalization properties of SVMs. We also propose improvements to computing posterior probability distributions from SVMs, and present a local-consistency potential measure that encourages spatial continuity. SVRFs can be e ciently trained, converge quickly during inference, and can be trivially augmented with kernel functions. SVRFs are more robust to class imbalance than Discriminative Random Fields (DRFs), and are more accurate near edges. Our results on synthetic data and a real-world tumor detection task show the superiority of SVRFs over both SVMs and DRFs.
Citation
C. Lee,
R. Greiner,
M. Schmidt.
"Support Vector Random Fields for Spatial Classification".
European Conference on Principles and Practice of Knowledge Discovery in Databases (PKDD), Porto, Portugal, pp 121-132, October 2005.
Keywords: |
machine learning, support vector machine, logistic regression, random field, SVRF, brain tumor |
Category: |
In Conference |
BibTeX
@incollection{Lee+al:PKDD05,
author = {Chi-Hoon Lee and Russ Greiner and Mark Schmidt},
title = {Support Vector Random Fields for Spatial Classification},
Pages = {121-132},
booktitle = {European Conference on Principles and Practice of Knowledge
Discovery in Databases (PKDD)},
year = 2005,
}
Last Updated: June 05, 2007
Submitted by Staurt H. Johnson