Conformal Prediction for Early Stopping in Mixed Integer Optimization

Stefan Clarke, Bartolomeo Stellato · 2026

Read Conformal Prediction for Early Stopping in Mixed Integer Optimization by Stefan Clarke, Bartolomeo Stellato (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-2602-01476.txt.

Description

Stefan Clarke and Bartolomeo Stellato use conformal prediction to stop mixed-integer solvers early. Solvers often find the best solution early yet spend most time proving it optimal. Their method learns, across similar cases, when to terminate a solver. A neural network reads the true optimality gap from solver state, and conformal prediction calibrates a stopping threshold with rigorous probabilistic quality guarantees. On five families from the distributional MIPLIB library, it cuts solve time by over 60% while keeping solutions within 0.1% of optimal with 95% probability.

Catalog record

CREATORStefan Clarke, Bartolomeo Stellato
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE34047
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602.01476

Browse the full library · About the Tower of Babel