1 Boosting We have seen so far how to solve classification (and other) problems when we have a data representation already chosen. We now talk about a procedure, known as boosting, which was originally discovered by Rob Schapire, and further developed by Schapire and Yoav Freund, that automatically