Split, Skip and Play: Variance-Reduced ProxSkip for Tomography Reconstruction is Extremely Fast

2026

Read Split, Skip and Play: Variance-Reduced ProxSkip for Tomography Reconstruction is Extremely Fast (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-2602-09527.txt.

Description

Iterative tomographic reconstruction pays twice per step: costly forward and adjoint projections for data fidelity, plus costly proximal computations often solved by inner iterations. Evangelos Papoutsellis and two co-authors study a method that cuts both bills at once in their 2026 paper. It couples randomized skipping of the proximal step with variance-reduced subset optimization of the data-fit term.

Trials on synthetic and real data show striking speedups of five to twenty times over non-skipped counterparts, the standard approach until now for efficiently solving these problems. By skipping expensive proximal work randomly while controlling variance in the data term, the method keeps accuracy while slashing compute. The authors present the work as groundwork for wider adoption of skipping methods across inverse problems.

Catalog record

YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE23 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-09527

Browse the full library · About the Tower of Babel