Call for NIPS 2008 Kernel Learning Workshop Submissions
Document Actions
J. Platt (2000)
Probabilities for SV Machines
In: Advances in Large Margin Classifiers, ed. by A.J. Smola and P.L. Bartlett and B. Schölkopf and D. Schuurmans, pp. 61-74, Cambridge, MA, MIT Press.
SVM do not immediately lead to confidence ratings. This problem is addressed by by fitting a logistic to the function values of a SVM in order to obtain Probabilities for SV Machines. The results are comparable to classical statistical techniques such as logistic regression while conserving the sparseness and thus numerical efficiency of SVMs. Pseudocode is given for easy implementation.