I worked out some thoughts about David McAllester's PAC-Bayes paper. lyx,tex, ps
jcl@cmu.edu