Variable Metric Implementations of the Proximal Point Algorithm

J. V. Burke

ABSTRACT. The Proximal Point Algorithm (PPA) is a method for solving inclusions of the form
0 lies in T(z) ,
where T is a monotone operator on a Hilbert space. The algorithm is one of the most powerful and versatile solution techniques for solving variational inequalities, convex programs, and convex-concave mini-max programs. In the context of convex programming, where the problem is to minimize a convex function f, the operator T is the convex subdifferential of f. In this case, the PPA can be viewed as an implementation of the method of steepest descent applied to the Moreau-Yosida regularization of f:
fr(x) =min{f(z) + (2r)-1 |z-x|2} .

Motivated by this interpretation, we consider a variable metric implementation of the method and study conditions under which such an implementation accelerates convergence. In this talk, we describe both the global and local convergence properties of the variable metric proximal point algorithm and provide conditions under which some standard matrix secant updating strategies yield the local super-linear convergence of the method.