Another way to say what you're proposing is to use a linear predictor, and to train it by globally optimizing 0-1 loss (just the number of mistakes that you make) on the full set of data that you have. Even ignoring computational issues (this can be shown to be NP-hard), you seem to be making several mistaken assumptions. I'd really recommend reading some basic stuff on generalization, but a couple of the mistaken assumptions are as follows:
1. That the particular input features that you've chosen are somehow the only possible choice. But who's to say that you shouldn't add new features which are the square of each original feature? Or maybe some cross product terms, like the product of the ith feature times the jth feature. Or maybe some good features to add would be the distance from each point you've seen so far. Etc. Continuing down this path, you basically get to the question discussed in the OP about choosing a kernel for SVMs. This is just one example where hyperparameters come into play, and you need some method for choosing them.
2. That a linear predictor is impervious to overfitting. Consider the extreme case (which comes up often) where you have millions or billions of features and far fewer examples (e.g., if features are n-gram occurrences in text, or gene expression data). Then it's likely that there are many settings of weights that fit the data perfectly, but there's no way to tell if you're just picking up on statistical noise, or if you've learned something that will make good predictions on new data that you encounter. In both theory and practice, you need some form of regularization, and along with this comes more hyperparameters, which need to be chosen.
Finally, by your reasoning, it seems like you would always choose a 1-nearest neighbors classifier [1] (because it will always end up with 0 error under the setting your propose). But there's no reason why this is in general a good idea.
Also, for people who are interested in the application of predicting college basketball with machine learning, there's a Google group that is worth joining:
As another commenter pointed out, the accuracy really needs to be evaluated using a validation set, not the test set--the approach described in the post is training with the testing data. In the field, we call this "cheating".
The basic idea of automatically tuning hyperparameters (the things this post discusses tuning with genetic algorithms) is cool, though, and is becoming a popular subject in machine learning research. A couple recent research papers on the topic are pretty readable:
Of particular note is the use of Theano for the machine learning heavy-lifting. If you do machine learning and haven't looked into Theano, you're probably making things harder on yourself than it needs to be.
Last year there was discussion where some HNers suggested that they'd be interested in participating in the competition if we allowed other forms of data:
http://news.ycombinator.com/item?id=2321009
This is the early announcement this year, where we're soliciting suggestions about other forms of data to include for this year.
I did research with Daphne (and co-authored a paper with her) in my senior year of undergrad, and she demanded a high standard of work, yes, but she was an excellent supervisor. Everybody knows how brilliant she is, but she also put a lot of effort into teaching my (also undergrad) project partner and I about how to formulate a research problem, how to do research, and how to present research. The primary concern appeared to be our personal growth, not the research machine (though that's not to say that the research wasn't important).
Working with her was one of the highlights of my undergrad education, and her class was great, too.
I'm personally having the same problem today, when the auto renewal date is a month away. I've filed two tickets without response. It seems the only solution people have found is to remove their credit card information from Google Checkout.
This was the intention. Around a month ago we started asking what data people would like to use. We incorporated some of that feedback to decide what data to use for this year.
If you have other suggestions, please let us know, and we'll add it for next year (if possible). The only thing we're trying to avoid is somebody coming in with a lot of data at the last minute, beyond the point when anybody else can realistically get it incorporated into their model.
One other issue that comes up is "garbage time". When a game isn't close (say in the last quarter of a blow-out), the stats are basically meaningless. Does Pomeroy have good ideas about how to deal with that?
This paper (referenced in the post) is also relevant. The application is to NBA basketball:
"Incorporating Side Information into Probabilistic Matrix Factorization Using Gaussian Processes." Ryan Prescott Adams, George E. Dahl, and Iain Murray. In Proceedings of the 26th Conference on Uncertainty in Artificial Intelligence, 2010.
There are many different actual tasks that technically are PASCAL challenges, but when people say "PASCAL VOC challenge" (Visual Object Classes), they typically mean either the _classification_ or _detection_ challenge:
Classification: For each of the twenty classes, predicting presence/absence of an example of that class in the test image.
Detection: Predicting the bounding box and label of each object from the twenty target classes in the test image.