Abstract | ||
---|---|---|
We propose practical recommendations for selecting metaparameters for SVM regression (that is, 驴 -insensitive zone and regularization parameter C). The proposed methodology advocates analytic parameter selection directly from the training data, rather than resampling approaches commonly used in SVM applications. Good generalization performance of the proposed parameter selection is demonstrated empirically using several lowdimensional and high-dimensional regression problems. In addition, we compare generalization performance of SVM regression (with proposed choice驴) with robust regression using 'least-modulus' loss function (驴 =0). These comparisons indicate superior generalization performance of SVM regression. |
Year | DOI | Venue |
---|---|---|
2002 | 10.1007/3-540-46084-5_112 | ICANN |
Keywords | Field | DocType |
support vector regression,proposed methodology advocate,proposed choice,high-dimensional regression problem,generalization performance,svm regression,good generalization performance,analytic parameter selection,svm application,proposed parameter selection,robust regression,loss function | Training set,Pattern recognition,Computer science,Regression analysis,Support vector machine,Polynomial regression,Svm regression,Robust regression,Regularization (mathematics),Artificial intelligence,Resampling,Machine learning | Conference |
Volume | ISSN | ISBN |
2415 | 0302-9743 | 3-540-44074-7 |
Citations | PageRank | References |
16 | 1.21 | 5 |
Authors | ||
2 |
Name | Order | Citations | PageRank |
---|---|---|---|
Vladimir Cherkassky | 1 | 1064 | 126.66 |
Yunqian Ma | 2 | 533 | 44.21 |