Applications of Statistical Field Theory in Deep Learning

Ringel, Zohar, Rubin, Noa, Mor, Edo, Helias, Moritz, Seroussi, Inbar · 2025

Read Applications of Statistical Field Theory in Deep Learning by Ringel, Zohar, Rubin, Noa, Mor, Edo, Helias, Moritz, Seroussi, Inbar (2025) 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-2502-18553.txt.

Introduction

Zohar Ringel and co-authors review how statistical field theory, a physics toolkit for wild probability landscapes, can explain deep learning. Neural networks learn functions drawn from vast distributions, which is exactly the terrain field theory was built to cross. The science of why networks generalize still lags their feats.

The review teaches the bridge step by step for newcomers. Field methods have already yielded sharp reads on three puzzles: why networks generalize to fresh data, why training leans toward simple fits, and how features form inside layers. Each result recasts a network question as a field average that can be computed. The authors map where the analogy holds and where it strains. Physicists gain a new playground, and learning theorists gain exact tools.

Catalog record

CREATORRingel, Zohar, Rubin, Noa, Mor, Edo, Helias, Moritz, Seroussi, Inbar
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE253 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2502-18553

Browse the full library · About the Tower of Babel