Multi-sphere support vector clustering

Authors:
Trung Le;Dat Tran;Phuoc Nguyen;Wanli Ma;Dharmendra Sharma
Affiliations:
Faculty of Information Sciences and Engineering, University of Canberra, ACT, Australia;Faculty of Information Sciences and Engineering, University of Canberra, ACT, Australia;Faculty of Information Sciences and Engineering, University of Canberra, ACT, Australia;Faculty of Information Sciences and Engineering, University of Canberra, ACT, Australia;Faculty of Information Sciences and Engineering, University of Canberra, ACT, Australia
Venue:
ICONIP'11 Proceedings of the 18th international conference on Neural Information Processing - Volume Part II
Year:
2011

Citing 5
Cited 0

The nature of statistical learning theory

The nature of statistical learning theory
Support vector clustering

The Journal of Machine Learning Research
Support Vector Data Description

Machine Learning
Data Mining: Practical Machine Learning Tools and Techniques, Second Edition (Morgan Kaufmann Series in Data Management Systems)

Data Mining: Practical Machine Learning Tools and Techniques, Second Edition (Morgan Kaufmann Series in Data Management Systems)
Multiple distribution data description learning algorithm for novelty detection

PAKDD'11 Proceedings of the 15th Pacific-Asia conference on Advances in knowledge discovery and data mining - Volume Part II

Quantified Score

Hi-index	0.00

Visualization

Abstract

Current support vector clustering method determines the smallest sphere that encloses the image of a dataset in feature space. This sphere when mapped back to data space will form a set of contours that can be interpreted as cluster boundaries for the dataset. However this method does not guarantee that the single sphere and the resulting cluster boundaries can best describe the dataset if there are some distinctive data distributions in this dataset. We propose multi-sphere support vector clustering to address this issue. Data points in data space are mapped to a high dimensional feature space and a set of smallest spheres that encloses the image of the dataset is determined. This set of spheres when mapped back to data space will form a set of contours that can be interpreted as cluster boundaries. Experiments on different datasets are performed to demonstrate that the proposed approach provides a better cluster analysis than the current support vector clustering method.