25 Years Of Conventional Evaluation Of Data Analysis Proves Worthless In Practice

Article was a thin on detail. What I take to be the actual paper http://www.signal.uu.se/Publications/pdf/p0804.pdf [Cross-validation and Bootstrapping are Unreliable in Small Sample Classification (PDF)] from http://www.signal.uu.se/Publications/ppapers.html [the author's list of journal publications].

Here is an explanation of cross-validation and bootstrapping: http://www.eecs.berkeley.edu/~pliang/cs294-spring08/lectures...

this methodology is worthless in practice when the total number of examples is small in relation to the natural variation that exists among different observations

... it therefore needs to be immediately replaces by Bayesian methods, for example, which can deliver reliable measures of the uncertainty that exists

This looks like dogmatic madness, but maybe the article just makes them sound bad. Can anyone comment on how reliable the uncertainty coming out of a Bayesian algorithms is?

I, for one, welcome our new Bayesian overlords....

Seriously, though-- it is precisely these kinds of applications that Bayesian methods are aimed at. Are they really that infrequently used in practice?