Skip to Main Content
A novel algorithm, robust kernel principal component analysis (robust KPCA), is proposed, based on research of the KPCA algorithm and its robustness. This algorithm generalizes the minimum error criteria of signal reconstruction to feature space, which can automatically recognize the outliers in the training sample set, and exterminates their effects on the accuracy of the KPCA algorithm via iterative computing. The robust KPCA algorithm not only retains the non-linearity property of KPCA, but has better robustness and improves the accuracy of KPCA. Simulation experiments show that the robust KPCA algorithm developed is better than the KPCA algorithm.
Date of Conference: 6-10 April 2003