Hidden geometry explains why kernel methods separate complex data so well

Are two sets of data genuinely different, or is it because of randomness? This question, known as the two-sample testing ...