Greedy GDA method for training data reduction and nonlinear feature extraction
Xiaowen Li · Kongzhi yu juece · 2011
Nonlinear feature extraction using standard generalized discriminant analysis(GDA) has high computational complexity in large datasets.Therefore,a greedy GDA(GGDA) is proposed to reduce training data and deal with the nonlinear feature extraction problem.Firstly,a subset is selected from the full training data by using the greedy technique of the greedy KPCA(GKPCA) method.Then,the feature extraction model is trained by using the GDA method with the subset instead of the full training data.Finally,classification experiments using data of several feature extraction methods are performed.The simulation results show that the feature extraction performance of both the GGDA and the GDA methods outperform that of other methods.In addition of retaining the performance of the GDA method,the GGDA method reduces the computational complexity of the nonlinear feature extraction in large datasets.