ResearchTrend.AI
  • Papers
  • Communities
  • Organizations
  • Events
  • Blog
  • Pricing
Papers
Communities
Social Events
Terms and Conditions
Pricing
Parameter LabParameter LabTwitterGitHubLinkedInBlueskyYoutube

© 2025 ResearchTrend.AI, All rights reserved.

  1. Home
  2. Papers
  3. 2505.15064
53
0

Generalization Through Growth: Hidden Dynamics Controls Depth Dependence

21 May 2025
Sho Sonoda
Yuka Hashimoto
Isao Ishikawa
Masahiro Ikeda
ArXiv (abs)PDFHTML
Main:26 Pages
Bibliography:3 Pages
Abstract

Recent theory has reduced the depth dependence of generalization bounds from exponential to polynomial and even depth-independent rates, yet these results remain tied to specific architectures and Euclidean inputs. We present a unified framework for arbitrary \blue{pseudo-metric} spaces in which a depth-\(k\) network is the composition of continuous hidden maps \(f:\mathcal{X}\to \mathcal{X}\) and an output map \(h:\mathcal{X}\to \mathbb{R}\). The resulting bound O((α+log⁡β(k))/n)O(\sqrt{(\alpha + \log \beta(k))/n})O((α+logβ(k))/n​) isolates the sole depth contribution in \(\beta(k)\), the word-ball growth of the semigroup generated by the hidden layers. By Gromov's theorem polynomial (resp. exponential) growth corresponds to virtually nilpotent (resp. expanding) dynamics, revealing a geometric dichotomy behind existing O(k)O(\sqrt{k})O(k​) (sublinear depth) and O~(1)\tilde{O}(1)O~(1) (depth-independent) rates. We further provide covering-number estimates showing that expanding dynamics yield an exponential parameter saving via compositional expressivity. Our results decouple specification from implementation, offering architecture-agnostic and dynamical-systems-aware guarantees applicable to modern deep-learning paradigms such as test-time inference and diffusion models.

View on arXiv
Comments on this paper