Moderating the outputs of support vector machine classifiers

被引:112
作者
Kwok, JTY [1 ]
机构
[1] Hong Kong Baptist Univ, Dept Comp Sci, Kowloon Tong, Peoples R China
来源
IEEE TRANSACTIONS ON NEURAL NETWORKS | 1999年 / 10卷 / 05期
关键词
Bayesian; evidence framework; moderated output; support vector machine;
D O I
10.1109/72.788642
中图分类号
TP18 [人工智能理论];
学科分类号
081104 ; 0812 ; 0835 ; 1405 ;
摘要
In this paper, we extend the use of moderated outputs to the support vector machine (SVM) by malting use of a relationship between SVM and the evidence framework. Tie moderated output is more in line with the Bayesian idea that the posterior weight distribution should be taken into account upon prediction, and it also alleviates the usual tendency of assigning overly high confidence to the estimated class memberships of the test patterns. Moreover, the moderated output derived here can be taken as an approximation to the posterior class probability. Hence, meaningful rejection thresholds can be assigned and outputs from several networks can be directly compared, Experimental results on both artificial and real-world data are also discussed.
引用
收藏
页码:1018 / 1031
页数:14
相关论文
共 44 条
[11]  
Cherkassky V.S., 1998, LEARNING DATA CONCEP, V1st ed.
[12]  
CORTES C, 1995, MACH LEARN, V20, P273, DOI 10.1023/A:1022627411411
[13]  
CRAVEN M, 1998, P 15 NAT C ART INT A
[14]  
Drucker H, 1997, ADV NEUR IN, V9, P155
[15]  
Dumais S., 1998, Proceedings of the 1998 ACM CIKM International Conference on Information and Knowledge Management, P148, DOI 10.1145/288627.288651
[16]  
FRAKES W, 1992, INFORMATION RETRIEVA
[17]  
HASTIE T, 1996, CLASSIFICATION PAIRW
[18]  
Joachims T., 1998, Lecture Notes in Computer Science, P137, DOI DOI 10.1007/BFB0026683
[19]  
KWOK JT, UNPUB IEEE T NEURAL
[20]  
Kwok JTY, 1998, ICONIP'98: THE FIFTH INTERNATIONAL CONFERENCE ON NEURAL INFORMATION PROCESSING JOINTLY WITH JNNS'98: THE 1998 ANNUAL CONFERENCE OF THE JAPANESE NEURAL NETWORK SOCIETY - PROCEEDINGS, VOLS 1-3, P347