Atelier Apprentissage 2006–2007
< précédent | suivant >
|Iterated Regularization for High-Dimensional Data: from Boosting to Twin Boosting|
Peter Bühlmann (ETH Zürich)
27 novembre 2006
Boosting algorithms have attracted much attention in the machine learning community as well as in related areas in statistics. They have proven to be very competitive in terms of prediction accuracy in a variety of applications. We present a statistical perspective on boosting, especially for high-dimensional data with many more covariates than sample size. We also show that Twin Boosting, an iterated boosting scheme yielding sparser solutions, can often improve upon boosting in terms of feature selection.