Qian, Yiming (2014) Self-tuning one-class support vector machines for data classification. Masters thesis, Memorial University of Newfoundland.
[English]
PDF
- Accepted Version
Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. Download (19MB) |
Abstract
Support Vector Machine (SVM) based classifiers are most popular models for data classification in machine learning. To obtain high classification accuracy, parameter tuning methods such as cross-validation are often applied, which is however timeconsuming. To address this problem, a simple, efficient and parameter-free algorithm is presented in this thesis. The algorithm is especially useful when dealing with datasets in the presence of label noise. Grown out of one-class SVM, the presented algorithm enjoys several distinct features: First, its decision boundary is learned based on both positive and negative examples, whereas the original one-class SVM training is only based on positive examples; Second, the internal parameters are self-tuned, which makes the algorithm handy to use even for first-time users. Compared with the benchmark method LIBSVM, the presented algorithm achieves comparable accuracy, while consuming only a fraction of the processing time. Applications in computer vision are presented to demonstrate the effectiveness of the algorithm.
Item Type: | Thesis (Masters) |
---|---|
URI: | http://research.library.mun.ca/id/eprint/8306 |
Item ID: | 8306 |
Additional Information: | Including bibliographical references (pages 58-66). |
Department(s): | Science, Faculty of > Computer Science |
Date: | June 2014 |
Date Type: | Submission |
Library of Congress Subject Heading: | Support vector machines; Computer algorithms; Text processing (Computer science); Pattern recognition systems; Computer vision |
Actions (login required)
View Item |