Adding PAM implementation

[lendle]

Example usage in combination with the Distance package: http://nbviewer.ipython.org/gist/lendle/8820029

[johnmyleswhite]

Very cool. What's the license for the materials you've based this on?

[lendle]

Good question. One reference looks like class notes (http://www.cs.umb.edu/cs738/pam1.pdf), and I do not see any license information on what appears to be the course website. The other two from wikipedia and wikibooks are have the Creative Commons Share Alike license.
I only used the descriptions of the algorithms in the references, not any code.

[johnmyleswhite]

If you only looked at descriptions, we're totally fine. It's only code that's troubling, since you can't copyright algorithms, only code.

[lendle]

Okay sounds good. I cited the first one in particular so it would be clear where my notation came from and because that formulation of the algorithm is not very intuitive.

By the way, are there plans at this point for a coherent clustering result type? I think that would make sense at last for partitioning methods.

[johnmyleswhite]

There aren't plans, but we'd be pretty open to hearing what you think we should do. Make things more generic would be a big gain for us.

[lendle]

I moved this comment to #13

[lindahua]

@lendle: may I suggest that you open a new issue specifically to discuss generic clustering interface?

It is better that this issue is for the PAM method.

[lendle]

I'm canceling this pull request so I can work on the PAM implementation a bit more, and because the package is getting moved to JuliaStats. I'll resubmit when I have a more mature implementation.