Dykstras algorithm with bregman projections: A convergence proof

From MaRDI portal





Dykstra's algorithm and the method of cyclic Bregman projections are often employed to solve best approximation and convex feasibility problems, which are fundamental in mathematics and the physical sciences. [\textit{Y. Censor} and \textit{S. Reich}, Commun. Appl. Anal. 2, 407-419 (1998; Zbl 0897.90155)] very recently suggested a synthesis of these methods, Dykstra's algorithm with Bregman projections, to tackle a nonorthogonal best approximation problem. They obtained convergence when each constraint is a halfspace. It is shown here that this new algorithm works for general closed convex constraints; this complements Censor and Reich's result and relates to a framework by [\textit{P. Tseng}, Math. Program. 59 A, 231-247 (1993; Zbl 0782.90073)]. The proof rests on original work [\textit{J. P. Boyle} and \textit{R. L. Dykstra}, Advances in Order Restricted Statistical Inference, Lect. Notes Stat. 37, 28-47 (1986; Zbl 0603.49024)] and on strong properties of Bregman distances corresponding to Legendre functions. Special cases and observations simplifying the implementation of the algorithm are also discussed.



Cites work


Cited in
(42)








This page was built for publication: Dykstras algorithm with bregman projections: A convergence proof

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q2709445)