Just to reinforce, these are the top "most influential" algorithms in machine learning. Most are no longer state-of-the-art.

Do you know what's state of the art then? I've touched on most of these & wouldn't mind learning more.

Ultimately, it depends on the kind of problem you have at hand and the data you're dealing with. The algorithms listed usu. will get you decent performance.

If you're dealing with large, sparse data this might serve you well: http://code.google.com/p/sofia-ml/

Some good stuff by Googlers.

Why is the main PDF linked through docs.google.com? What's the added value of this? I can only see privacy drawbacks, but I'm a bit paranoid on this side, so maybe I'll stuck with the "it's just not needed" point :-)

Direct PDF clicky: http://www.cs.uvm.edu/~icdm/algorithms/10Algorithms-08.pdf

Though letting Google deal with PDFs (including potential malware) has its upside too ...

I'm really surprised Learning Classifier Systems don't receive more attention. I thought I was well versed in the subject - is there something I'm missing, anyone?