The support vector machine can be approximated geometrically by enclosing the data of the classes by convex hulls. Then you can e.g. using the Rotating Calipers, you can search the widest band/border between these convex hulls and thus you have the maximum minimum distance between the two classes and can make predictions. That would be a purely geometric approach to solving the Support Vector machine. But would that solution still be classified as supervised learning or is it only a part of computational geometry?

Thanks in advance.

Are algorithms that solve the k-center problem part of machine learning or computational geometry?


1 Answer 1


I'm not clear on the details of your geometric algorithm, but if it works, then it would be the application of computational geometry to a supervised learning problem.

It's not the case that "ideas" are divided up into mutually exclusive categories such that an idea can be part of "computational geometry" or "supervised learning" but not both. After all, it is humans who define the categories.

  • $\begingroup$ Thank you very much for your detailed answer, shimao. In other words, one would not leave the field of machine learning/supervised learning through this geometric approach? $\endgroup$
    – Code Now
    Commented Nov 8, 2018 at 5:21

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.