An O(1/κ) convergence rate for the variable stepsize bregman operator splitting algorithm
Document Type
Article
Publication Date
1-1-2016
Abstract
An earlier paper proved the convergence of a variable stepsize Bregman operator splitting algorithm (BOSVS) for minimizing φ(Bu) + H(u), where H and φ are convex functions, and φ is possibly nonsmooth. The algorithm was shown to be relatively efficient when applied to partially parallel magnetic resonance image reconstruction problems. In this paper, the convergence rate of BOSVS is analyzed. When H(u) = ∥ Au -f∥2, where A is a matrix, it is shown that for an ergodic approximation uκ obtained by averaging κ BOSVS iterates, the error in the objective value φ(Buκ)+H(uκ) is O(1/κ). When the optimization problem has a unique solution u∗, we obtain the estimate ∥uκ -u∗∥ = O(1/ √κ). The theoretical analysis is compared to observed convergence rates for partially parallel magnetic resonance image reconstruction problems where A is a large dense ill-conditioned matrix.
Publication Source (Journal or Book title)
SIAM Journal on Numerical Analysis
First Page
1535
Last Page
1556
Recommended Citation
Hager, W., Yashtini, M., & Zhang, H. (2016). An O(1/κ) convergence rate for the variable stepsize bregman operator splitting algorithm. SIAM Journal on Numerical Analysis, 54 (3), 1535-1556. https://doi.org/10.1137/15100401X