I'm working on a specific classifier which can classify the below toy data, which is a 2-class problem but the classes are distributed among each others. So it is not possible to solve it by typical SVM or Knn or prototype based classifiers.
But in order to show the practical use of my method, i need some real dataset generally similar to the above condition. It can be n-class problem, where the class data are coordinated among each others, but locally separable, and shouldn't be possible to achieve high performance using Knn or SVM classifiers.
I was wondering if such dataset is available anywhere?