论文信息 - Breaking SVM Complexity with Cross-Training

Breaking SVM Complexity with Cross-Training

We propose to selectively remove examples from the training set using probabilistic estimates related to editing algorithms (Devijver and Kittler, 1982). This heuristic procedure aims at creating a separable distribution of training examples with minimal impact on the position of the decision boundary. It breaks the linear dependency between the number of SVs and the number of training examples, and sharply reduces the complexity of SVMs during both the training and prediction stages.

[1] Tom Downs,et al. Exact Simplification of Support Vector Solutions , 2002, J. Mach. Learn. Res..

[2] Alexander J. Smola,et al. Learning with kernels , 1998 .

[3] Thorsten Joachims,et al. Making large scale SVM learning practical , 1998 .

[4] Chih-Jen Lin,et al. A tutorial on?-support vector machines , 2005 .

[5] C. G. Hilborn,et al. The Condensed Nearest Neighbor Rule , 1967 .

[6] Vladimir N. Vapnik,et al. The Nature of Statistical Learning Theory , 2000, Statistics for Engineering and Information Science.

[7] Christopher J. C. Burges,et al. Simplified Support Vector Decision Rules , 1996, ICML.

[8] Ingo Steinwart,et al. Sparseness of Support Vector Machines---Some Asymptotically Sharp Bounds , 2003, NIPS.

[9] Peter E. Hart,et al. The condensed nearest neighbor rule (Corresp.) , 1968, IEEE Trans. Inf. Theory.

[10] Dennis L. Wilson,et al. Asymptotic Properties of Nearest Neighbor Rules Using Edited Data , 1972, IEEE Trans. Syst. Man Cybern..

[11] Thomas G. Dietterich,et al. Editors. Advances in Neural Information Processing Systems , 2002 .

[12] Igor Durdanovic,et al. Parallel Support Vector Machines: The Cascade SVM , 2004, NIPS.