Fast Compute via MC Boosting

Sarah Polson, Vadim Sokolov · 2026

Read Fast Compute via MC Boosting by Sarah Polson, Vadim Sokolov (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2602-05032.txt.

Description

Sarah Polson and Vadim Sokolov survey a faster way to solve the linear systems that sit inside modern learning pipelines in this 2026 statistics paper. Training and inference loops solve such systems again and again, and exact solvers get expensive when only a few components of the answer are actually needed.

Their topic is Monte Carlo boosting: random-walk estimators that refine an approximate solution step by step. The paper unifies several variants in one notation, covering Neumann-series forms, forward and adjoint estimators, and sequential correction schemes, with extensions to least-squares problems and iterative methods used in data augmentation. Benchmarks compare the Monte Carlo family against Jacobi and Gauss-Seidel iteration. The results map out where random methods win, framing them as a practical compute primitive.

Catalog record

CREATORSarah Polson, Vadim Sokolov
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE31 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-05032

Browse the full library · About the Tower of Babel