Skip to content
DraconiticResearch
Search Sources API
197,768 studies31 sourcesUpdated just now
PreprintarXiv licencearXiv

A Full Adam Theorem for Spectral Heavy-Tail Onset

Abstract

We prove a full Adam theorem for spectral heavy-tail onset in a closed Gaussian Stein-Hermite teacher-student state-evolution model. The theorem begins with the actual full-batch Adam recurrences, derives the population gradient by Stein-Hermite calculus, proves finite-width covariance concentration, converts multi-step Adam momentum into an exact non-centered Gaussian sign kernel, controls the diagonal Adam denominator by a basis-homogenization theorem, derives a regularly varying projected update response from a Hermite edge-transfer theorem, pushes the response through the exact Gram update, and proves approximate-target KL contraction with matching upper and lower hitting bounds. The final law is (τ_\varepsilon=Θ(Δ_1^{-γ}d^ρ\log(Ψ_0/\varepsilon))), where (Δ_1) is the first spike-bulk spectral gap. The result is full in the following precise sense: every step from Adam's momentum and denominator to the spectral hitting law is formalized inside the closed state-evolution model. We also prove that a stronger arbitrary-gradient Adam theorem is impossible, and that exact two-step linear-network loss dynamics do not identify factor spectra or heavy-tail hitting times.

Related

  1. Local limit Theorem for Directed polymers in a random environment in the full weak disorder regime
    Stefan Junk, Hubert Lacoin
    arXivPreprintarXiv licenceFull textProbabilityarXiv
  2. Spectral aspects of random heavy-tailed tensors
    Remi Bonnin, Alexis Imbert
    arXivPreprintarXiv licenceFull textProbabilityarXiv
  3. THE OPTIMAL EIGENVALUE COUNT IN AUBRY'S SPECTRAL SPHERE THEOREM FOR DIMENSIONS AT LEAST FOUR
    Anass Nifa
    HALPreprintCC BY-NC-ND 4.0Full textGeometry & topologyHAL
  4. A general theorem in spectral extremal graph theory
    John Byrne, Dheer Noal Desai, Michael Tait
    Transactions of the American Mathematical SocietyPeer-reviewedLicence not statedFull text24 citedCombinatorics & discrete mathematicsGeometry & topologyOpenAlex
  5. Gradient estimates for the Green kernel under spectral Ricci bounds, and the stable Bernstein theorem in ℝ⁴
    Xavier Cabré, Giovanni Catino, Luciano Mari +2
    Transactions of the American Mathematical SocietyPeer-reviewedLicence not statedFull text10 citedGeneral mathematicsOpenAlex
BibTeX
@misc{liu2026full,
  title = {{A Full Adam Theorem for Spectral Heavy-Tail Onset}},
  author = {Zongmin Liu},
  howpublished = {arXiv},
  year = {2026},
  month = sep,
  doi = {10.48550/arxiv.2609.12996},
  eprint = {2609.12996},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2609.12996},
}
RIS
TY  - UNPB
TI  - A Full Adam Theorem for Spectral Heavy-Tail Onset
AU  - Zongmin Liu
PY  - 2026
DA  - 2026/09/01
PB  - arXiv
DO  - 10.48550/arxiv.2609.12996
UR  - https://arxiv.org/abs/2609.12996
AB  - We prove a full Adam theorem for spectral heavy-tail onset in a closed Gaussian Stein-Hermite teacher-student state-evolution model. The theorem begins with the actual full-batch Adam recurrences, derives the population gradient by Stein-Hermite calculus, proves finite-width covariance concentration, converts multi-step Adam momentum into an exact non-centered Gaussian sign kernel, controls the diagonal Adam denominator by a basis-homogenization theorem, derives a regularly varying projected update response from a Hermite edge-transfer theorem, pushes the response through the exact Gram update, and proves approximate-target KL contraction with matching upper and lower hitting bounds. The final law is (τ_\varepsilon=Θ(Δ_1^{-γ}d^ρ\log(Ψ_0/\varepsilon))), where (Δ_1) is the first spike-bulk spectral gap. The result is full in the following precise sense: every step from Adam's momentum and denominator to the spectral hitting law is formalized inside the closed state-evolution model. We also prove that a stronger arbitrary-gradient Adam theorem is impossible, and that exact two-step linear-network loss dynamics do not identify factor spectra or heavy-tail hitting times.
ER  -