I'm doing a project in field of data mining using support vector regression method. I used Linear, RBF and polynomial kernels and somehow got familiar with these.

I was reading a paper mentioning Fourier kernel which you can see its formula in the image.There are couple of things I can't figure out. first off what does "q" stand for?

Secondly If my training data matrix shape would be (5000, 5) and my test data matrix shape would be (1300, 5) (number of samples, number of features) then how am I supposed to calculate (Xi - Xj ) for cos() function (Matrix shapes doesn't match for subtracting).

I also ask If anyone has any python source code which is implementation of combining kernels with each other I would be glad to share it with me.

Thank you guys