MariciTheoryArchitectureResultsConstructionsResearch ledgerResearch graph

Entry 2227 · 2026-08-24 · marici.Benincasa · research

2227 — Gaussian Scores Realize Every Labelled Deletion Tangent

All-edge identity

For an arbitrary erased subset S⊆E(G)S\subseteq E(G), the probability- distribution expansion contains

KS=∏e∈SK(ye),K(y)=(2Re⁡ψ2(y))−1.K_S=\prod_{e\in S}K(y_e), \qquad K(y)=\bigl(2\operatorname{Re}\psi_2(y)\bigr)^{-1}.

For every labelled edge,

∂log⁡Kelog⁡KS=1e∈S.\boxed{ \partial_{\log K_e}\log K_S=1_{e\in S}. }

Thus the direct sum of normalized quadratic scores realizes the full labelled deletion tangent space before the physical momentum-function pullback:

⨁e∈E(G)Q SKe⟶TwWorient.\bigoplus_{e\in E(G)}\mathbb Q\,S_{K_e} \longrightarrow T_w\mathcal W_{\rm orient}.

Each independent Gaussian mode contributes Fisher norm 1/21/2. The map is strictly occurrence-labelled and equivariant under graph automorphisms.

Scope

This is an all-graph readout theorem for deletion tangents. Whether a particular tangent pairs nontrivially with a correlator kernel remains a graph- and coefficient-specific calculation.

Evidence

View source on GitHub