Tobias Schlottmann
2007-Jul-02 15:06 UTC
[R] Implementing Support Vector Regression by ipop in kernlab
Dear R users, To implement the Support Vector Regression by a general QP, I'm using for example the function ipop in kernlab package. Let me please have some questions for my understanding. In the example for ipop function, the matrix A includes only variable y. However, in Support Vector Regression formulation, we have these constraints: y-f(x)+b<=e f(x)+b-y<=e How vector y does correspond to the matrix A in this formulation? How can we include the constant b of above constraints in ipop? In the first term of objective function (according to the kernlab package document), variable x should correspond to slack variables by means of SVR formulation. However, in the rest of the objective function (quadratic term) , variable x corresponds to w in SVR formulation. Therefore, as far as I understand x corresponds to two variables. Or what is my missing? I thank you so much for your guids. I'm not an expert in this area. Please help me to understand ipop. with the best regards, T. S. --------------------------------- Building a website is a piece of cake. [[alternative HTML version deleted]]