Globally convergent BFGS method for nonsmooth convex optimization (Q1573986)

For the unconstrained minimization of a nonsmooth strongly convex function \(f:\mathbb{R}^n\to \mathbb{R}\), the authors apply the BFGS method to its Moreau-Yosida regularization. The algorithm they propose makes use of approximate values of the regularized function and of its gradient; it is proved to converge to the unique optimal solution.