COLT
2001
Springer
14 years 4 months ago
2001
Springer
We consider geometric conditions on a labeled data set which guarantee that boosting algorithms work well when linear classifiers are used as weak learners. We start by providing ...