Calcolo ( IF 1.7 ) Pub Date : 2021-03-19 , DOI: 10.1007/s10092-021-00406-9 Kui Du , Xiao-Hui Sun
The randomized Kaczmarz algorithm has received considerable attention recently because of its simplicity, speed, and the ability to approximately solve large-scale linear systems of equations. In this paper we propose randomized double and triple Kaczmarz algorithms to solve extended normal equations of the form \(\mathbf{A}^\top \mathbf{Ax}=\mathbf{A}^\top \mathbf{b}-\mathbf{c}\). The proposed algorithms avoid forming \(\mathbf{A}^\top \mathbf{A}\) explicitly and work for arbitrary \(\mathbf{A}\in \mathbb {R}^{m\times n}\) (full rank or rank-deficient, \(m\ge n\) or \(m<n\)). Tight upper bounds showing exponential convergence in the mean square sense of the proposed algorithms are presented and numerical experiments are given to illustrate the theoretical results.
中文翻译:
随机的双重和三重Kaczmarz用于求解扩展正规方程
随机化的Kaczmarz算法最近由于其简单性,速度和近似求解大型线性方程组的能力而备受关注。在本文中,我们提出了随机的双重和三重Kaczmarz算法来求解形式为\(\ mathbf {A} ^ \ top \ mathbf {Ax} = \ mathbf {A} ^ \ top \ mathbf {b}-\\ mathbf {c} \)。所提出的算法避免显式地形成\(\ mathbf {A} ^ \ top \ mathbf {A} \)并为\ mathbb {R} ^ {m \ times n} \中的任意 \(\ mathbf {A} \(满等级或等级不足,\(m \ ge n \)或\(m <n \))。紧的 给出了算法均方指数收敛的上限,并进行了数值实验验证了理论结果。