Iterative regularization for convex regularizers

We study iterative/implicit regularization for linear models, when the bias is convex but not necessarily strongly convex. We char- acterize the stability properties of a primal- dual gradient based approach, analyzing its convergence in the presence of worst case deterministic noise. As a main example, we specialize and illustrate the results for the problem of robust sparse recovery. Key to our analysis is a combination of ideas from regularization theory and optimiza- tion in the presence of errors. Theoreti- cal results are complemented by experiments showing that state-of-the-art performances can be achieved with considerable computa- tional speed-ups.