Nonlinear Canonical Correlation Analysis for Discrimination Based on Kernel Methods
Ping Bo Sun · Chinese Journal of Computers · 2004
In this paper, we generalize the Canonical Correlation Analysis (CCA) for discrimination to yield a new nonlinear learning machine by using kernel methods. We name it as Kernel Canonical Correlation Discriminant Analysis (KCCDA), which is a powerful technique for extracting nonlinear features from high-dimensional data sets. To overcome the problems of computation complexity, we also proposed a adaptive learning algorithm for KCCDA based on online sparsification. Extensive experiments on artificial and real-world data sets demonstrate the competitiveness of KCCDA and our adaptive learning algorithm. Finally, from the theoretical viewpoint we proved that KCCDA is identical to the Kernel Fisher Discriminant analysis(KFD) except for an unimportant scale factor.