GENERALIZED BREGMAN PROJECTIONS IN CONVEX FEASIBILITY PROBLEMS

We present a method for finding common points of finitely many closed convex sets in Euclidean space. The Bregman extension of the classical method of cyclic orthogonal projections employs nonorthogonal project ions induced by a convex Bregman function, whereas the Bauschke and Bo rvein method uses Bregman/Legendre functions. Our method works with ge neralized Bregman functions (B-functions) and inexact projections, whi ch are easier to compute than the exact ones employed in other methods . We also discuss subgradient algorithms with Bregman projections.