NeurIPS 2019
Sun Dec 8th through Sat the 14th, 2019 at Vancouver Convention Center
Paper ID:6662
Title:Comparing distributions: $\ell_1$ geometry improves kernel two-sample testing


		
The paper shows L_p distance of kernel mean embedding as a distance measure of distributions, and shows that L_1-based statistics has higher power than the standard L2 distance. The paper is clearly written and technically sound. The proposed method will be beneficial in the field of kernel-based distance measures and statistical test.