Fast Adaboost Training Algorithm by Dynamic Weight Trimming

Evolutionary extreme learning machine based on dynamic Adaboost ensemble

“…But, it is very difficult to give the upper bound of weighted error on weak learning in advance. So application of Boosting was limited [15] . In this paper we propose an evolutionary extreme learning machine based on dynamic Adaboost ensemble in order to overcome mentioned faults.…”

“…Fuzzy activation function provides some advantages such as low computational burden and easy implementation in hardware. And dynamic Adaboost ensemble [15] is new ensemble learning algorithm based on Adaboost [18] and SWTAdaboost [19] . The ensemble learning algorithm is called Dynamic Weight Trimming Adaboost (DWTAdaboost).…”

“…It had effectively broken such limitations as weak generalization capability of single network model. Besides, Boosting has two fatal limitations: one is that it gives in advance the upper bound of weighted error on weak learning algorithm [15] ; the other one is that it is overdependent on data and weak learning machine, and it is too sensitive to data noising [16] .…”

The Machine Knows What You Are Hiding: An Automatic Micro-expression Recognition System

“…However, due ired by MI, the training of Gentleboost even would beco nd the idea of dynamical weight trimming (DWT) [18] for this problem. This implementation is still called DW samples are filtered according to a threshold whic weight distribution over training data at the correspond with weight less than the threshold are not used for train s not trained on the whole training samples and is selec ed error of selected classifier may even be more than or eq non-effective classifiers, is dynamically chan on error.…”

“…As shown in Table 1, the accuracy of Gentleboost with DWT was only slightly lower than the original Gentleboost. The larger value of can bring faster training speed, but it will also introduce more errors [18]. A trade-off between speed and accuracy must be considered.…”