Hi Joao.
There is no build-in way to achieve this.
You can create the kernel matrix of the tensor product of the linear and
the nonlinear kernel explicitly and pass that to SVR.
That will allow the SVR to use both kernels at once.
I think it should just be the product of the two kernel values.
Hi!
I have a question regarding support vectors regression.
Is it possible to develope a kernel that combines the linear and the rbf
kernel, that is, it uses both kernels and applies each one to a region
where each kernel fits better the data?
I thought of this because I was asked to fit this sim
