Features

The ALvsPK challenge (http://www.agnostic.inf.ethz.ch/) is prolonged until August 1st, 2007.

The participants may win one of 6 prizes: best overall entry in the "agnostic learning track" and best entry for each dataset in the "prior knowledge track".

Initial results (http://clopinet.com/isabelle/Projects/agnostic/ALPK-Results-Mar-07.ppt)

indicate that you can get better results by building your domain knowledge into your preprocessing or your prediction algorithm, than if you use provided features and a generic learning machine. But is this really true? Try to compete in your own domain of expertise: marketing, handwriting recognition, drug discovery/QSAR, text classification, ecology.