Just a quick idea.
Take an example of two classes, label red and blue. I wonder if you can make a classifier with the help of image filters rather than defining parameters to time consuming more analytic classifiers like SVM (RBF).
Further. Taking inspiration from the gaussian filterered image. I wonder if classfiers can be built with holes, islands and unsure areas. Islands being a red cluster within the blue group. An unsure area being the area with a much blue and red dots near each other.