The Algorithm of Nonlinear Feature Extraction for Large scale Data set

In order to deal with the storage and computation problem for large-scale data set, an efficient iterative method is proposed. Some mathematical tricks are firstly used to transform the kernel matrix and the sample vector can be gotten in feature space. Then, the sample vector is used for iterative algorithm to extract nonlinear discriminant vectors. The space complexity of proposed method reduces from 2 ( ) m  to ( ) m  . The experimental results validate the effectiveness of the proposed method.