Conference Proceedings

A Modified ν-SV Method For Simplified Regression

A Shilton, M Palaniswami

Proceedings of International Conference on Intelligent Sensing and Information Processing Icisip 2004 | IEEE | Published : 2004

Abstract

In the present papeiw e describea new algorithm for Support Vector regression (SVR). Like standard ℏ-SVR algorithms, this algorithm automatically adjusts the radius of insensitivity (tube width ε) to fit the data. However, this is achieved without additional complexity in the optimisation problem. Moreover, by careful modification of the kernel function, we are able to significantly simplify the form of the dual SVR optimisation problem.

University of Melbourne Researchers