Variable quasi-Bregman monotone sequences (Q501975)

This paper unifies the work of \textit{P. L. Combettes} and \textit{B. C. Vũ} [Nonlinear Anal., Theory Methods Appl., Ser. A, Theory Methods 78, 17--31 (2013; Zbl 1266.65087)], where the notion of quasi-Fej\'{er} sequences is extended to the situation that allows the metric to vary with iterations; and that of \textit{H. H. Bauschke} et al. [SIAM J. Control Optim. 42, No. 2, 596--636 (2003; Zbl 1049.90053)], which introduces the notion of Bregman monotonicity. This is accomplished via introducing the so-called quasi-Bregman monotone sequence and its associated asymptotic properties. These results are applied to a couple of optimization algorithms in Banach spaces, with extensive details. In particular, a variable Bregman proximal point algorithm in Banach spaces is introduced and a convex feasibility problem in Banach spaces is treated.