DBM - Hinton & Salakhutdinov paper

[carloderamo]

Pylearn2 implementation of the whole procedure to train a DBM described by Hinton & Salakhutdinov in "An Efficient Learning Procedure For Deep Boltzmann Machines".
This pr adds a fast version for the test script to be run on travis together with some changes to the format and the readme file. I also moved augment_input from datasets to /scripts/dbm.

[dwf]

@carloderamo Can you rebase and squash this into a single commit?

As this is all concerning DBM stuff that nobody knows anything about other than @goodfeli I'm going to punt the review to him.

[goodfeli]

You should not need a new Dataset object to do this.

[goodfeli]

Since I've requested relatively major changes based on the small parts I have reviewed, I'm going to quit reviewing early and will review the rest if / when the first changes are made.

[carloderamo]

Currently I have a lot of work to be done for completing my MSc thesis on
May. I'll work on my pull request, making another one with all the changes
requested, as soon as I'll finish it.
Best regards.

Carlo D'Eramo

Since I've requested relatively major changes based on the small parts I
have reviewed, I'm going to quit reviewing early and will review the rest
if / when the first changes are made.

—
Reply to this email directly or view it on GitHub
#1239 (comment).