Sequential Preconditioned Conjugate Gradient Method for Linear Statistical Models

DOI10.48550/arXiv.2607.25272MaRDI QIDQ7284450arXivFDO

Authors Guan-Yu Chen, Dong-Yue Xie, Xi Yang, Zun-Hao Zheng

Publication date 28 July 2026

Published in arXiv

Tldr We develop a randomized iterative method for least-squares estimation in linear statistical models, called Sequential Preconditioned Conjugate Gradient Method (SPCG), which constructs sketched subproblems to accelerate convergence. SPCG achieves faster convergence by exploiting warm-starts and sketch sizes that increase with iterations, reducing computational cost while maintaining OLS accuracy.

Full work available at URL https://doi.org/10.48550/arXiv.2607.25272

Mathematics Subject Classification ID

No records found.

arXiv classification

math.NAstat.ML

Cites work

• A statistical perspective on randomized sketching for ordinary least-squares
• Asymptotics for Sketching in Least Squares Regression
• Blendenpik: Supercharging LAPACK's Least-Squares Solver
• Iterative Hessian sketch: fast and accurate solution approximation for constrained least-squares
• LSRN: A parallel iterative solver for strongly over- or underdetermined systems
• Matrix Computations
• Numerical Optimization
• Randomized Numerical Linear Algebra : A Perspective on the Field With an Eye to Software
• Randomized Numerical Linear Algebra: Foundations & Algorithms

This page was built for publication: Sequential Preconditioned Conjugate Gradient Method for Linear Statistical Models

Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q7284450)