A ROBUST BOOSTING METHOD FOR MISLABELED DATA
Natsuki Sano, Hideo Suzuki, Masato Koda · Journal of the Operations Research Society of Japan · 2004
We propose a new, robust boosting method by using a sigmoidal function as a loss function. In deriving the method, the stagewise additive modelling methodology is blended with the gradient descent algorithms. Based on intensive numerical experiments, we show that the proposed method is actually better than AdaBoost and other regularized method in test error rates in the case of noisy, mislabeled situation.