Besides, I'll appreciate if one's introduce me an useful tutorial on SVM. I do't need mathematical tutorial or practical tutorial. I'm looking for general concepts. A tutorial with graphical explanation (as my question) would be appreciated.
Thanks in advance.
Is this some kind of homework? Some additional information would be nice, for example I assume that all the points in the images are related to a particular class.
The decision boundary of a SVM with linear kernel is an hyperplane, hence image (A). Instead, the decision boundary of a SVM with Gaussian kernel is identified by a weighted sum of Gaussians centered in the support vectors, hence it is expected to be rather "smooth", probably images (C) and (E). Which one is associated to a particular parameter depends whether in your case the parameter is in the usual form 1/2sigma or simply at the numerator. There is a very nice Quora answer on the Guassian kernel:
http://www.quora.com/Support-Vector-Machines/What-is-the-intuition-behind-Gaussian-kernel-in-SVM
As for the general tutorial, there is a known paper in Nature Biotechnology which skips over all the math:
[1] Noble, William S. "What is a support vector machine?." Nature biotechnology 24.12 (2006): 1565-1567.
You may find some interesting explanations there.
Thanks Simone for your reply and your beneficial link. This question is't a homework. I'm just practicing machine learning and pattern recognition and I ran into this question.
- Badges
- Science topic