Adaptive sparse kernel principal component analysis for computation and store space constrained-based feature extraction

On the training samples number and kernel function and its parameter en- dured by Kernel Principal Component Analysis, this paper presents one-class support vector based Adaptive Sparse Kernel Principal Component Analysis (ASKPCA) through reducing the training samples with sparse learning based least squares support vector ma chine and adaptive self-optimizing kernel structure according to the input training sam ples. ASKPCA is adaptive to the computation and store space constrained applications, such as small size of hardware platform based image retrieval, medical assistant diag nosis system, and so on. Moreover, the few meaningful samples are found with solving the constraint optimization equation, and these training samples are used to compute the kernel matrix which decreases the computing time and saving space, and the algorithm is to increase the recognition accuracy and computing efficiency in the image process ing under the hardware computing platform on the limit training samples. Experimental results on UCI datasets, ORL and YALE face databases and Wisconsin Breast Cancer database show that it is feasible to improve KPCA and Sparse Kernel Principal Compo nent Analysis (SKPCA) on saving consuming space and optimizing kernel structure.