The goal of mmdproject is to learn about and present maximum mean discrepancy in machine learning. It is a method - and test - for comparing two distributions. I still don’t know how it compares with Kolmogorov Smirnov, but I anticipate that the 2012 JMLR piece discusses this point.