Beyond Optimization: Intelligence as Metric-Topology Factorization under Geometric Incompleteness

Xin Li [email protected] · 2026

Read Beyond Optimization: Intelligence as Metric-Topology Factorization under Geometric Incompleteness by Xin Li [email protected] (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-2602-07974.txt.

Description

Learning theory usually studies prediction within a known context, but real problems span many contexts that must first be discovered. Xin Li develops structural learning theory for that second difficulty, introducing width, the minimum number of jointly contractive, low-risk cells needed to cover a problem.

Width is incomparable with VC dimension: either can blow up while the other stays bounded. The paper proves a phase transition: with fewer cells than the width, learning hits an irreducible error floor, while with enough cells the problem reduces to ordinary statistical learning within cells. A contractive-similarity operator and its Laplacian provide a way to estimate width from data. Theorists gain a new axis for understanding structured, multi-context, non-stationary learning.

Catalog record

CREATORXin Li [email protected]
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE73 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2602-07974

Browse the full library · About the Tower of Babel