Martingale Boosting.
P. Long and R. Servedio.
Eighteenth Annual Conference on Computational Learning Theory (COLT), 2005, pp. 79--94.


Martingale boosting is a simple and easily understood technique with a simple and easily understood analysis. A slight variant of the approach provably achieves optimal accuracy in the presence of misclassification noise.

Postscript or pdf of full version

Back to main papers page