The Schwarzian Bridge in a Single Sigmoid Neuron: Exact Loewner Determinants and Gaussian Saturation Anisotropy
Admitted after checks by named AI models and editor approval. Not peer reviewed. Screening record
We identify a local coefficient identity linking two geometries of an increasing activation g at a centered inflection: for h_p=(g')^p under Gaussian input, the initial radial-tangential anisotropy is 1-p Sg(0)r^2+O(r^4), while the adjacent-point Loewner determinant has coefficient g'(0)^2 Sg(0)/6. For the logistic sigmoid we give a closed determinant formula, a dimension-minimal 2 by 2 PSD-order witness certified with 256-bit interval arithmetic, profile-generic finite-radius curvature bounds, exact large-saturation constants, and strict anisotropy monotonicity for Bernoulli and squared-output profiles. Classical Schwarzian/composition results and prior one-unit Fisher decompositions are explicitly separated from the new bridge and exact refinements. The optimization consequence is restricted to stationary fixed-step local descent; no global, deep-network, or non-Gaussian claim is made.
Preview PDF on this page v1
Uses your browser's PDF viewer. Open the PDF directly if a preview is unavailable.
Cite this paper
Lluis Eriksson (2026). The Schwarzian Bridge in a Single Sigmoid Neuron: Exact Loewner Determinants and Gaussian Saturation Anisotropy. AIRR.SCIENCE, ARR-2026-53CTRKDSP685PT51, v1. https://airr.science/papers/ARR-2026-53CTRKDSP685PT51/versions/v1/
Use this version link to cite the manuscript you read. A Working paper label means the version has not completed AIRR screening. AIRR screening is not journal peer review. Import BibTeX, RIS or CSL JSON into your reference manager to apply a citation style.
Version history
The paper identifier remains stable. Each version has its own immutable release, timestamp and version identifier.
Verification record
- Frontier-model screening
- Not assessed
- Source integrity
- Pass
- Bibliographic integrity
- Partial
- Reproducibility
- Partial
- Lean 4
- Not applicable
Recorded under ARR-VERIFY-1.0. AIRR verification and screening are not peer review.
AI assistance statement
OpenAI Codex assisted with corpus and literature triage, theorem discovery, adversarial proof and novelty audits, numerical replay, drafting, and artifact preparation. The author reviewed the claims and remains responsible for the paper.
Frontier-model screening
Status: not_assessed. Any listed reports correspond to this exact version under ARR-SCREEN-1.0; no absent assessment is represented as a pass.
Model assessments and review context
No supported exact-version report is published for this exact version. Missing evidence is not scored as zero.
No model reports are published for this version.
A model assessment is not peer review or a correctness certificate. AIRR preserves disagreement, exact-version provenance and later reassessments.
Editorial disclosure
Founder-owned pilot record: the author and AIRR founder-editor are the same person. Acceptance records a technically valid deposit after hostile audits, not independent peer review or a guarantee that the paper is correct.