I'm an applied mathematician working on approximation theory and its applications to optics. I've been working with a Gaussian Kernel Density Estimator (KDE) for some while, but figured out I should be looking broader, and understand better what kernel I should be working with and how.

I'm looking for a shorter-then-text book introduction to KDE's, either as a short review, book chapter or even a slideshow. The main questions I need to be adressed:

  1. What is the main criteria for choosing a specific KDE method over the other?
  2. I understand the meaning of the "window size", but what are the main methods for choosing it?
  • 1
    $\begingroup$ Does this help stats.stackexchange.com/q/244012/35989 ? $\endgroup$
    – Tim
    Nov 9, 2017 at 20:37
  • $\begingroup$ @Tim it helps immensly, but it also helped me to sharpen my question. See new edit $\endgroup$
    – Amir Sagiv
    Nov 10, 2017 at 7:26
  • 1
    $\begingroup$ (1) What do you mean by "specific KDE method"? (2) There is lots of methods, it was answered here: stats.stackexchange.com/questions/168/… $\endgroup$
    – Tim
    Nov 10, 2017 at 8:51
  • $\begingroup$ By (1) I mean, what specific kernel function to use, i.e., Gaussian, Bspline etc. @Tim $\endgroup$
    – Amir Sagiv
    Nov 11, 2017 at 14:57
  • $\begingroup$ Bspline is not a kernel and the choice of the kernel in general does not matter for vast majority of the cases. $\endgroup$
    – Tim
    Nov 11, 2017 at 15:43


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.