Continuous Optimization Seminar - Kay Barshad

Friday, May 26, 2023 1:30 pm - 1:30 pm EDT (GMT -04:00)

Title: Generalized Modular String Averaged Procedure and Its Applications to Iterative Methods

Speaker: Kay Barshad
Affiliation: University of Waterloo
Location: MC 6029

Abstract: A modular string averaging procedure (MSA, for short) for a finite number of operators  was first introduced by Reich and Zalas in 2016. The MSA concept provides a flexible  algorithmic framework for solving various feasibility problems such as common fixed point and convex feasibility problems. In 2001 Bauschke and Combettes introduced  the notion of coherence and applied it to proving weak and strong convergence of many iterative methods. In 2019 Barshad, Reich and Zalas proposed a stronger variant of coherence which provides a more convenient sufficient convergence condition for such methods. In this paper we combine the ideas of both modular string averaging and coherence. Focusing on extending the above MSA procedure to an infinite sequence of operators with admissible controls, we establish strong coherence of its output operators. Various applications of these concepts are presented with respect to weak and strong 13 convergence. They also provide important generalizations of known results, where the weak convergence of sequences of operators generated by the MSA procedure with  intermittent controls was considered.