TLDR: The references in the docstrings have weird/wrong/very different formats. "There should be one-- and preferably only one --obvious way to do it."

I was looking at the reference links of some classifiers when I realized they were not formated correctly. When I was about to submit a pull request I realized that many references have a different format and according to the Zen of Python

There should be one-- and preferably only one --obvious way to do it.

Some references

• are referencable, some not.
• are not wrapped correctly
• have links to the PDFs, some don't, some links are dead.

I think references should be referencable, be as concise as possible (authors, title, conference/book, etc). I'm not sure about the link to the PDFs because the URIs change over time and lead to dead links. What do you think?

My suggestion:

   References
    ----------

    .. [1] Hinton, G. E., Osindero, S. and Teh, Y. A fast learning algorithm
           for deep belief nets. Neural Computation 18, pp 1527-1554.

I'd be willing to spend a few hours on the weekend to fix the references once we decide on how it's supposed to look like.

Also, #3912 should document the format for the references.