EcoSta 2022: Start Registration
View Submission - EcoSta2022
A0364
Title: Total stability of SVMs and localized SVMs Authors:  Hannes Koehler - University of Bayreuth (Germany) [presenting]
Andreas Christmann - University of Bayreuth (Germany)
Abstract: Regularized kernel-based methods such as support vector machines (SVMs) typically depend on the underlying probability measure (respectively data set) as well as on the regularization parameter and the kernel that are used. Whereas classical statistical robustness only considers the effect of small perturbations in the probability measure alone, we investigate how the resulting predictor is influenced by simultaneous slight variations in the whole triple of probability measure, regularization parameter and kernel. Existing results from the literature are considerably generalized and improved. In order to also make them applicable to big data, where regular SVMs suffer from their super-linear computational requirements, the results are transferred to the context of localized learning.