Title :
Classifier combination based on active learning
Author :
Yi, Xing ; Kou, Zhongbao ; Zhang, Changshui
Author_Institution :
Dept. of Autom., Tsinghua Univ., Beijing, China
Abstract :
In this paper, we propose classifier combination based on active learning, which deals with the design of classifier combination systems as training a combiner at the aggregation level and introduces SVM active learning into the design of this multi-category decision combiner. This algorithm greatly reduces the number of labeled data the classifier system needs in order to achieve satisfactory performance. This algorithm consists of two main steps: firstly, designing and training first level classifiers which can output posterior probability vectors as the input of the second level combiner, secondly, designing second level combiner based on SVM active learning and classifying testing samples with this combiner. Experiments on standard database show that our algorithm performs better than current classifier combination rules when considering both labeling cost and classification accuracy.
Keywords :
learning (artificial intelligence); maximum likelihood estimation; pattern classification; probability; support vector machines; SVM active learning; aggregation level; classifier combination rules; multicategory decision combiner; posterior probability vectors; second level combiner; Algorithm design and analysis; Costs; Databases; Labeling; Laboratories; Learning systems; Machine learning; Pattern recognition; Support vector machine classification; Support vector machines;
Conference_Titel :
Pattern Recognition, 2004. ICPR 2004. Proceedings of the 17th International Conference on
Print_ISBN :
0-7695-2128-2
DOI :
10.1109/ICPR.2004.1334054