01:49:13	Hossein Rashidi:	Please enable subtitles
02:12:13	maurizio:	Sorry, my daughter unmuted
03:09:58	Bob Cousins:	Jerry Friedman at PhyStat 2003, https://inspirehep.net/files/345e70f422e5306b7540c80ed1dbcff3  "Thus, the power of these tests can be highly sensitive to the learning machine employed. Particular choices depend on the types of potential differences between the distributions that are deemed most important to detect. For example, if the distributions tend to be different on a large fraction of the variables, near{neighbor or kernel methods will provide high power. On the other hand if they tend to differ on only a relatively small number of variables, decision trees will provide greater sensitivity."