Efficient KPCA-Based Feature Extraction: A Novel Algorithm and Experiments
KPCA has been widely used for feature extraction. It is noticeable that the efficiency of KPCA-based feature extraction is in inverse proportion to the size of the training sample set. In order to speed up KPCA-based feature extraction, we develop a novel algorithm(i.e. IKPCA) which improves KPCA with a distinctive viewpoint. The algorithm is methodologically consistent with KPCA with clear physical meaning. Experiments on several benchmark datasets illustrate that IKPCA-based feature extraction is much faster than KPCA-based feature extraction. The ratio of IKPCA-based feature extraction time to KPCA-based feature extraction time may be smaller than 0.30. Furthermore, the classification accuracy corresponding to IKPCA is comparable with KPCA.
KeywordsPrincipal Component Analysis Feature Extraction Feature Space Training Sample Kernel Principal Component Analysis
Unable to display preview. Download preview PDF.
- Turk, M., Pentland, A.: Face Recognition Using Eigenfaces, Proc. IEEE Conf. On Computer Vision and Pattern Recognition, (1991)586–591Google Scholar
- Bian, Z., Zhang, X.: Pattern Recognition (in Chinese), Tsinghua University Press, Beijing, (2000)Google Scholar
- Jin, Z., Davoine, F., Lou, Z., Yang, J.-Y.: A Novel PCA-based Bayes Classifier and Face Analysis, IAPR International Conference on Biometrics (ICB2006), Hong Kong, Jan. 5–7, (2006)Google Scholar
- Scholkopf, B., Smola, A., Muller, K.R.: Kernel Principal Component Analysis, Artificial Neural Networks-ICANN’97, Berlin, (1997) 583–588Google Scholar
- Xu, Y., Yang, J.-Y., Lu, J.-F.: An Efficient Kernel-based Nonlinear Regression Method for Two-class Classification, Proceedings of 2005 International Conference on Machine Learning and Cybernetics, Guangzhou, China, August, (2005) 4442–4445Google Scholar