Neural Nets

Fascinating talk on the state of the Art of Neural Nets. I signed up for Geoff Hinton's Coursera course, but haven't had a chance to go through it yet. At first glance It seemed kind of dry, but in this video he has a sharp sense of humor.
http://www.youtube.com/watch?v=vShMxxqtDDs

Idea: Sports Mega-nerd

I had an idea this weekend:
Hanging out with my old college buddies (that are significantly more in the Sports Info Demographic than I), we inevitably ended up listening to some sports radio, or watching some ESPN sports talk show. It occurred to me that all of these talking heads were essentially just going on and on about their various sports opinions, and occasionally dropping a few minor win/loss stats to back it up. My idea was in essence that this whole industry is ripe for someone to take this to the next level and (in real time) stitch together a story based on predictive models from various sources of data. Basically what a vegas odds maker does, but with somewhat less rigor, less consequences of failure, and more light-hearted humor.
In essence I see this as an extension of what Nate Silver did for politics. Before Nate you had a select few "Opinionated Experts", after Nate you had the power of a statistical model.

Algebraic Classifiers

This could be interesting… If only I had time to wade through it. 

"Well, it turns out that the bayesian classifier has the algebraic structure of a monoid, a group, and a vector space.  HLearn uses a new cross-validation algorithm that can exploit these algebraic structures.  The standard algorithm runs in time , where  is the number of "folds" and  is the number of data points.  The algebraic algorithms, however, run in time .  In other words, it doesn't matter how many folds we do, the run time is constant!"

http://izbicki.me/blog/hlearn-cross-validates-400x-faster-than-weka

http://izbicki.me/public/papers/icml2013-algebraic-classifiers.pdf