SVM Kerkel Scale Auto

Hello!

I am currently working with SVMs. I am using the Gaussian kernel function and I want to quickly find a good parameter for my training data (I am only concerned with the sigma/gamma, not with the soft margin C. For the soft margin I am using another method)

In the Matlab documentation is says: "Pass the data to fitcsvm, and set the name-value pair arguments 'KernelScale','auto'. Suppose that the trained SVM model is called SVMModel. The software uses a heuristic procedure to select the kernel scale. The heuristic procedure uses subsampling. Therefore, to reproduce results, set a random number seed using rng before training the classifier."

So, what I wanted to know: is this heuristic trying to select the "best" parameter with respect to my training data?

Thank you,

[sh10101]

Hi,

I am also looking for an answer to you question. Did you find a suitable answer during your time studying SVMs?

Thank you,