Use individual kernel functions in Matlab's SVMtrain

I got two questions concerning the use of the SVM function built into Matlab (2011a)

My ‚Training’ matrix consists of more than two columns, thus I’m dealing with a multidimensional problem. Can the quadprog (‚QP’) algorithm (Matlab Optimization toolbox is installed) deal with that?

And my second question concerns the use of individual kernel functions. I don’t get it to work, can you please help me to correct the code?

I use the following code to create a linear kernel for testing purpose (I know that it is already the standard, but this may be a good starting point):

SVMStruct = svmtrain(Training,Group,'autoscale',true,'method','QP','kernel_function',@kfun);

function K = kfun(U, V)
K = u*v;

Answer 1

[Léon]

May somebody please help me? Thank you very much!

Answer 2

[Walter Roberson]

SVM in current versions of MATLAB is strictly limited to two class problems.