Generalization performance of support vector classifiers for density level detection

Hong Chen, Yicong Zhou, Yi Tang, Yuan Yan Tang, Zhibin Pan

Abstract: This paper investigates the generalization performance of support vector classifiers for density level detection (DLD) when the input term belongs to a separable Hilbert space. The estimate of learning rate for DLD problem is established by Rademacher average and iterative techniques, which is independent of the assumption of covering number used in the previous literature.