On adaptive stochastic extended iterative methods for solving least squares

TL;DR

This paper introduces an adaptive stochastic extended iterative method for solving least squares problems, improving upon the randomized extended Kaczmarz method with proven linear convergence and demonstrated numerical advantages.

Contribution

The paper develops a new adaptive stochastic iterative method with theoretical convergence guarantees, extending existing randomized methods for least squares solutions.

Findings

Proves R-linear convergence in expectation.

Demonstrates numerical superiority over non-adaptive methods.

Addresses an open problem in adaptive stochastic optimization.

Abstract

In this paper, we propose a novel adaptive stochastic extended iterative method, which can be viewed as an improved extension of the randomized extended Kaczmarz (REK) method, for finding the unique minimum Euclidean norm least-squares solution of a given linear system. In particular, we introduce three equivalent stochastic reformulations of the linear least-squares problem: stochastic unconstrained and constrained optimization problems, and the stochastic multiobjective optimization problem. We then alternately employ the adaptive variants of the stochastic heavy ball momentum (SHBM) method, which utilize iterative information to update the parameters, to solve the stochastic reformulations. We prove that our method converges $R$-linearly in expectation, addressing an open problem in the literature related to designing theoretically supported adaptive SHBM methods. Numerical experiments show that our adaptive stochastic extended iterative method has strong advantages over the non-adaptive one.