Are there any good resources regarding how to design kernels for regression problems, specifically time-series regression type of problem. I am finding the choice of a kernel for regression extremely unintuitive.

For a classification problem with hinge loss, the idea is that the kernel should map similar examples closely while keep dis-similar examples apart. What is the equivalent notion for the ILF loss used in SVR?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.