Many fundamental machine learning tasks can be formulated as a problem of with vector-valued functions, where we learn multiple scalar-valued functions together. Although there is some generalization analysis on different specific algorithms under the empirical risk minimization principle, unifying regularization framework still lacking. In this paper, initiate regularized by presenting bounds ...