Hi all,
I am currently using the built-in "fitcsvm" function to train a classifier and I am slightly confused by the documentation. I would like to compare the performance of linear and RBF kernels, which is easy enough. However, when I wanted to tune parameters, I could not find an obvious way of setting the RBF kernel sigma value. There is the "KernelScale" parameter but I am confused as to what it actually does as it can be set for linear and polynomial kernels as well?
I am missing something obvious in the documentation?
Thanks for your help.
Nicolas