ENTROPY-LIKE PROXIMAL METHODS IN CONVEX-PROGRAMMING

We study an extension of the proximal method for convex programming, w here the quadratic regularization kernel is substituted by a class of convex statistical distances, called phi p-divergences, which are typi cally entropy-like in farm. After establishing several basic propertie s of these quasi-distances, we present a convergence analysis of the r esulting entropy-like proximal algorithm. Applying this algorithm to t he dual of a convex program, we recover a wide class of nonquadratic m ultiplier methods and prove their convergence.