Title :
A new Bayesian design method for support vector classification
Author :
Chu, Wei ; Keerthi, S. Sathiya ; Ong, Chong Jin
Author_Institution :
Dept. of Mech. Eng., Nat. Univ. of Singapore, Singapore
Abstract :
We apply popular Bayesian techniques on a support vector classifier. We propose a novel differentiable loss function called trigonometric loss function with the desirable characteristic of natural normalization in the likelihood function, and then follow standard Gaussian processes techniques to set up a Bayesian framework. In this framework, Bayesian inference is used to implement model adaptation, while keeping the merits of support vector classifier, such as sparseness and convex programming. Moreover, we put forward class probability in making predictions. Experimental results on benchmark data sets indicate the usefulness of this approach.
Keywords :
Bayes methods; Gaussian processes; convex programming; inference mechanisms; pattern classification; support vector machines; Bayesian design method; Bayesian inference; Bayesian techniques; benchmark data sets; class probability; convex programming; differentiable loss function; likelihood function; model adaptation; natural normalization; sparseness; standard Gaussian process techniques; support vector classification; trigonometric loss function; Adaptation model; Bayesian methods; Computer networks; Design methodology; Gaussian processes; High performance computing; Mechanical engineering; Static VAr compensators; Supervised learning; Vectors;
Conference_Titel :
Neural Information Processing, 2002. ICONIP '02. Proceedings of the 9th International Conference on
Print_ISBN :
981-04-7524-1
DOI :
10.1109/ICONIP.2002.1198189