Oganeditse A. Boikanyo

A proximal point algorithm converging strongly for general errors

In this paper a proximal point algorithm (PPA) for maximal monotone operators with appropriate regularization parameters is considered. A strong convergence result for PPA is stated and proved under the general condition that the error sequence tends to zero in norm. Note that Rockafellar (SIAM J Control Optim 14:877–898, 1976) assumed summability for the error sequence to derive weak convergence of PPA in its initial form, and this restrictive condition on errors has been extensively used so far for different versions of PPA. Thus this Note provides a solution to a long standing open problem and in particular offers new possibilities towards the approximation of the minimum points of convex functionals.

The Method of Alternating Resolvents Revisited

The purpose of this article is to prove a strong convergence result associated with a generalization of the method of alternating resolvents introduced by the authors in convergence of the method of alternating resolvents [4] under minimal assumptions on the control parameters involved. Thus, this article represents a significant improvement of the article mentioned above.

Proximal Point Methods Revisited

The proximal point methods have been widely used in the last decades to approximate the solutions of nonlinear equations associated with monotone operators. Inspired by the iterative procedure defined by B. Martinet (1970), R.T. Rockafellar introduced in 1976 the so‐called proximal point algorithm (PPA) for a general maximal monotone operator. The sequence generated by this iterative method is weakly convergent under appropriate conditions, but not necessarily strongly convergent, as proved by O. Guler (1991). This fact explains the introduction of different modified versions of the PPA which generate strongly convergent sequences under appropriate conditions, including the contraction‐PPA defined by H.K. Xu in 2002. Here we discuss Xu’s modified PPA as well as some of its generalizations. Special attention is paid to the computational errors, in particular the original Rockafellar summability assumption is replaced by the condition that the error sequence converges to zero strongly.