First-order methods for regularization — General first-order methods use only the values of objective/constraints and their subgradients to optimize non-smooth (including constrained) function [5]. Such methods that are oblivious to