Abstract
We present Mixture of Support Vector Data Descriptions (mSVDD) for one-class classification or novelty detection. A mixture of optimal hyperspheres is automatically discovered to describe data. The model consists of two parts: log likelihood to control the fit of data to model (empirical risk) and regularization quantizer to control the generalization ability of model (general risk). Expectation Maximization (EM) principle is employed to train the model. We demonstrate the advantage of the proposed model: if learning mSVDD in the input space, it simulates learning single hypersphere in the feature space and the accuracy is thus comparable but the training time is significantly shorter.
Original language | English |
---|---|
Title of host publication | Proceedings of 2015 2nd National Foundation for Science and Technology Development Conference on Information and Computer Science |
Editors | Nguyen Quoc Dinh, Tran Xuan Tu |
Place of Publication | Piscataway NJ USA |
Publisher | IEEE, Institute of Electrical and Electronics Engineers |
Pages | 135-140 |
Number of pages | 6 |
ISBN (Electronic) | 9781467366403, 9781467366380 |
ISBN (Print) | 9781467366397 |
DOIs | |
Publication status | Published - 2015 |
Externally published | Yes |
Event | National Foundation for Science and Technology Development Conference on Information and Computer Science 2015 - Ho Chi Minh, Vietnam Duration: 16 Sept 2015 → 18 Sept 2015 Conference number: 1st https://edas.info/web/nics14/index.html https://web.archive.org/web/20150807190107/http://nafosted-nics.org/ |
Conference
Conference | National Foundation for Science and Technology Development Conference on Information and Computer Science 2015 |
---|---|
Abbreviated title | NICS 2015 |
Country/Territory | Vietnam |
City | Ho Chi Minh |
Period | 16/09/15 → 18/09/15 |
Internet address |
Keywords
- kernel method
- mixture model
- Mixture of experts
- one-class classification