“…It can be seen that BDLDA is also more robust than the two other classifiers. The reason is that BDLDA benefit from robustness of LDA (Bertels, Jacques, Neuberg, & Gatot, 1999) and generalization property of Adaboost scheme (Murua, 2002). Moreover, regarding the high dimensionality of the feature vectors (286), BDLDA uses Direct LDA as its weak learner which reduces the dimension of input vectors while it increases the discrimination of the transformed feature vectors (Gao & Davis, 2006).…”