Advanced Search
DSpace@MIT

Comparing Support Vector Machines with Gaussian Kernels to Radial Basis Function Classifiers

Research and Teaching Output of the MIT Community

Show simple item record

dc.contributor.author Schoelkopf, B. en_US
dc.contributor.author Sung, K. en_US
dc.contributor.author Burges, C. en_US
dc.contributor.author Girosi, F. en_US
dc.contributor.author Niyogi, P. en_US
dc.contributor.author Poggio, T. en_US
dc.contributor.author Vapnik, V. en_US
dc.date.accessioned 2004-10-20T20:48:54Z
dc.date.available 2004-10-20T20:48:54Z
dc.date.issued 1996-12-01 en_US
dc.identifier.other AIM-1599 en_US
dc.identifier.other CBCL-142 en_US
dc.identifier.uri http://hdl.handle.net/1721.1/7180
dc.description.abstract The Support Vector (SV) machine is a novel type of learning machine, based on statistical learning theory, which contains polynomial classifiers, neural networks, and radial basis function (RBF) networks as special cases. In the RBF case, the SV algorithm automatically determines centers, weights and threshold such as to minimize an upper bound on the expected test error. The present study is devoted to an experimental comparison of these machines with a classical approach, where the centers are determined by $k$--means clustering and the weights are found using error backpropagation. We consider three machines, namely a classical RBF machine, an SV machine with Gaussian kernel, and a hybrid system with the centers determined by the SV method and the weights trained by error backpropagation. Our results show that on the US postal service database of handwritten digits, the SV machine achieves the highest test accuracy, followed by the hybrid approach. The SV approach is thus not only theoretically well--founded, but also superior in a practical application. en_US
dc.format.extent 6 p. en_US
dc.format.extent 2032389 bytes
dc.format.extent 277809 bytes
dc.format.mimetype application/postscript
dc.format.mimetype application/pdf
dc.language.iso en_US
dc.relation.ispartofseries AIM-1599 en_US
dc.relation.ispartofseries CBCL-142 en_US
dc.subject AI en_US
dc.subject MIT en_US
dc.subject Artificial Intelligence en_US
dc.subject radial basis function networks en_US
dc.subject support vector machines en_US
dc.subject pattern recognition en_US
dc.subject machine learning en_US
dc.subject VC-dimension en_US
dc.subject performance comparison en_US
dc.subject model selection en_US
dc.title Comparing Support Vector Machines with Gaussian Kernels to Radial Basis Function Classifiers en_US


Files in this item

Name Size Format Description
AIM-1599.ps 1.938Mb Postscript
AIM-1599.pdf 271.2Kb PDF

This item appears in the following Collection(s)

Show simple item record

MIT-Mirage