Hypothesis testing

We propose an E-values based classifier two-sample test that has much stronger finite-sample type-I error control than existing methods.