Fast adaboost algorithm based on weight constraints

This paper presents a fast Adaboost algorithm based on weight constraints, which can shorten the training time when dealing with a larger training data set. In the algorithm, the sample feature space is divided into several groups. Considering that the number of groups and sample dimension feature space are closely related, when the feature space dimension exceeds the number set group, the sample according to the set threshold array searches. Meanwhile, considering the sample weights update related to the number of misclassification samples, sample weights update adds to the number of misclassification, limiting its next iteration of the weights, thereby effectively preventing excessive weight adaptation phenomenon. Experiment shows that the algorithm achieves better results on UCI datasets.