Classifier geometrical characteristic comparison and its application in classifier selection

Authors:
Hui Zhu;Xianglong Tang
Affiliations:
Department of Computer Science and Engineering, Pattern Recognition Research Center, Harbin Institute of Technology, 92 West Dazhi Street, P.O. Box 352, Harbin 150001, China;Department of Computer Science and Engineering, Pattern Recognition Research Center, Harbin Institute of Technology, 92 West Dazhi Street, P.O. Box 352, Harbin 150001, China
Venue:
Pattern Recognition Letters
Year:
2005

Citing 11
Cited 1

Small Sample Size Effects in Statistical Pattern Recognition: Recommendations for Practitioners

IEEE Transactions on Pattern Analysis and Machine Intelligence
Network generalization differences quantified

Neural Networks
The Random Subspace Method for Constructing Decision Forests

IEEE Transactions on Pattern Analysis and Machine Intelligence
A Combination Fingerprint Classifier

IEEE Transactions on Pattern Analysis and Machine Intelligence - Graph Algorithms and Computer Vision
Complexity Measures of Supervised Classification Problems

IEEE Transactions on Pattern Analysis and Machine Intelligence
Feature Extraction Based on Decision Boundaries

IEEE Transactions on Pattern Analysis and Machine Intelligence
Measures of Diversity in Classifier Ensembles and Their Relationship with the Ensemble Accuracy

Machine Learning
Data Complexity Analysis for Classifier Combination

MCS '01 Proceedings of the Second International Workshop on Multiple Classifier Systems
Pretopological Approach for Supervised Learning

ICPR '96 Proceedings of the International Conference on Pattern Recognition (ICPR '96) Volume IV-Volume 7472 - Volume 7472
On the Nonlinearity of Pattern Classifiers

ICPR '96 Proceedings of the International Conference on Pattern Recognition (ICPR '96) Volume IV-Volume 7472 - Volume 7472
Engineering multiversion neural-net systems

Neural Computation

Naive random subspace ensemble with linear classifiers for real-time classification of fMRI data

Pattern Recognition

Quantified Score

Hi-index	0.10

Visualization

Abstract

Classifier selection is an important step in designing multiple classifier systems. In this paper classifier geometrical characteristic comparison methods including volume occupation difference comparison and decision boundary curvature difference comparison are proposed. Based on the comparison which is directly carried out on training samples, difference between classifiers can be measured. The usefulness of these methods to measure the difference between classifiers' decision space and their influence on recognition rate of multiple classifier systems are confirmed by a series of experiments. Then a classifier selection strategy based on these comparison methods is introduced. The experiment results show that it is a useful selection strategy for selecting classifiers to combine.