Reshaping Global Loop Structure to Accelerate Local Optimization by Smoothing Rugged Landscapes Preprint: APS/123-QED

Timothée Leleu, Sam Reifenstein, Atsushi Yamamura, Surya Ganguli · 2026

Read Reshaping Global Loop Structure to Accelerate Local Optimization by Smoothing Rugged Landscapes Preprint: APS/123-QED by Timothée Leleu, Sam Reifenstein, Atsushi Yamamura, Surya Ganguli (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-2602-01490.txt.

Description

Timothee Leleu and co-authors reshape global loop structure to speed local optimization. Frustrated graphical models trap optimization in rugged landscapes shaped by local and loop structure. Bethe approximation views the graph as a tree, ignoring it. The M-layer construction copies the graph M times, reconnecting edges randomly, from the original graph to Bethe at large M. The team generalizes it with a mixing kernel Q setting layer connection probabilities, shaping loops, and layer fluctuations decay, raising the odds of reaching the global minimum. Costs fall across Ising benchmarks.

Catalog record

CREATORTimothée Leleu, Sam Reifenstein, Atsushi Yamamura, Surya Ganguli
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE122236
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602.01490

Browse the full library · About the Tower of Babel