Outline
- Abstract
- Keywords
- 1. Introduction
- 2. Multi-Category Svm
- 3. Uncertainty Sampling
- 4. Uncertainty Sampling-Based Msvm
- 5. Experimental Results
- 6. Conclusions
- References
رئوس مطالب
- چکیده
- کلیدواژه ها
- 1. مقدمه
- 2. SVM چند دسته ای
- 3. نمونه برداری عدم قطعیت
- 4. MSVM مبتنی بر نمونه برداری عدم قطعیت
- 5. نتایج آزمایشی
- 6. نتیجه گیری ها
Abstract
Among the SVM-based methods for multi-category classification, “1-a-r”, pairwise and DAGSVM are most widely used. The deficiency of “1-a-r” is long training time and unclassifiable region; the deficiency of pairwise and DAGSVM is the redundancy of sub-classifiers. We propose an uncertainty sampling-based multi-category SVM in this paper. In the new method, some necessary sub-classifiers instead of all N × (N − 1)/2 sub-classifiers are selected to be trained and the uncertainty sampling strategy is used to decide which samples should be selected in each training round. This uncertainty sampling-based method is proved to be accurate and efficient by experimental results on the benchmark data.
Keywords: Multi-category classification - Pairwise - SVM - Uncertainty samplingConclusions
In this paper, we propose a novel uncertainty sampling based method, US_MSVM, to solve multi-category classification problems. In the new method, sub-classifiers are trained in order of their significances and those unhelpful sub-classifiers are ignored. The uncertainty sampling strategy is used to decide which samples should be trained in the next round. When testing, the final result is the integrative opinion of all trained sub-classifiers.
Experimental results on real-world data set show that, Precision and Recall of US_MSVM are comparable to those of pairwise in the condition that the training round is much less than that of the pairwise. The US_MSVM can be used as a substituted version of pairwise.