A1110
Title: Lp-consistency of regularized kernel methods and its connection to risk consistency
Authors: Hannes Koehler - University of Bayreuth (Germany) [presenting]
Abstract: It is well-known that risk consistency is a property that is satisfied by regularized kernel methods such as support vector machines under mild conditions. The close connection of risk consistency to $L_p$-consistency is investigated and established for a considerably wider class of loss functions than has been done before. From this, it is derived that the examined regularized kernel methods are indeed $L_p$-consistent as well. The attempt to transfer this result to shifted loss functions surprisingly reveals that this shift does not reduce the assumptions needed on the underlying probability measure to the same extent as it does for many other results regarding regularized kernel methods.
