A modified /spl nu/-SV method for simplified regression

In the present paper we describe a new algorithm for support vector regression (SVR). Like standard /spl nu/-SVR algorithms, this algorithm automatically adjusts the radius of insensitivity (tube width /spl epsiv/) 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.