jghrng

EDIT: Continued to read the actual article -- I didn't know about the mammography madness in the US. AFAIK doing X-Rays or CTs without reasonable suspicion is considered personal injury here (in Germany).

Well you can always simply ask the author -- if he has got the copyright, he can grant you permission.

For some problems, it may be nice to have a generative model as offered by RBMS (although Rifai et al. published a sampling method for contractive auto-encoders recently: http://icml.cc/2012/papers/910.pdf). I feel like with RBMs, you can design models which incorporate prior knowledge more "easily" (you may end up with pretty complex models...), e.g. the conditional RBM, the mean-covariance RBM or the spike & slab RBM. Additionally, there's the deep boltzmann machine that consists of multiple layers that are jointly trained in an RBM-like fashion.

Auto-encoders are straightforward to understand and implement. With contractive terms or denoising, the are powerful feature extractors as well.

But as you already noted, if you "just" want to have a good classifier, I think it pretty much boils down to personal preference since you're going to spend some effort on making these techniques work well on your problem anyway.

EDIT: Try the US version: http://translate.google.com.mx/?hl=en#auto/en/filled%20with

It should be more useful (but more expensive) to lighten up the actual crosswalk signs on the road as soon as pedestrians are approaching.

Great work!