A revised AdaBoost algorithm: FM-AdaBoost
Yanfeng Zhang, HE Pei-kun · 2010
In view of ensemble equivalence, this paper proposes a revised AdaBoost algorithm: FM-AdaBoost. It can ensure the ensemble error rates are the least by F-module, which filter classifiers after all of the iteration finish. At the same time, with the optional M-module it can ensure the training error rates decreases monotonously, which improves the training velocity effectively. In the end, simulation results show the algorithm is valid.