Kernelmethodsareamongthemostpopulartechniquesinmachinelearning.Fromaregularizationperspec-tivetheyplayacentralroleinregularizationtheoryastheyprovideanaturalchoiceforthehypothesesspaceandtheregularizationfunctionalthroughthenotionofreproducingkernelHilbertspaces.Fromaprobabilisticper-