Deterministic AdaBoost algorithm based on FLDF
Jong Chan Lee, Jun Wu, Keon Myung Lee · 2008
AdaBoost is an algorithm with a procedure of selecting the data events from a dataset at each iteration sequence. The data events are selected stochastically using a random number generator. In this paper, a deterministic AdaBoost algorithm is proposed in contrast to the usual stochastic one. For doing this we derive the modified Fisherpsilas formulas moderated to the deterministic method. These formulas contain a scheme to treat data set with weight vector. To verify the performance of proposed algorithm, we compare with the results of different measurements with the deterministic and the stochastic method, by gradually increasing the prune rate and the number of weak learner in the network structure. Through the result of these experiments, we show that our proposed method has higher performance than typical stochastic one.