If (like me) you didn't know what to make of this, the video of the talk by Raymond Hettinger is here:
http://www.viddler.com/explore/python-italia/videos/85/
He starts with this example at ~9:30 in, and boy is it cool...
to be frank Neural Nets are out of fashion in both CS Research as well as Data Mining Practice.
Try Algorithmic Game Theory, Probabilistic Algorithms, Probabilistic Graphical Models, Bayesian Nets, Support Vector Machines etc.
Why are they out of fashion? Is it because it is too hard to figure out why a particular neural net makes certain decisions? Are they generally too unwieldy and unpredictable?
I'm very interested in neural nets but I don't know much about alternative tools/models, so I'm keen to hear your reasoning.
Bing's relevance and ranking engine is supposedly based upon neural nets.