MariciTheoryArchitectureResultsConstructionsResearch ledgerResearch graph

Entry 2214 · 2026-08-24 · marici.Benincasa · research

2214 — The Gaussian Kernel Tangent Has a Normalized Quadratic Score

One-mode calculation

For the normalized Gaussian boundary measure

Pa(x)=aπ e−ax2,K=⟨x2⟩=12a,P_a(x)=\sqrt{\frac a\pi}\,e^{-ax^2}, \qquad K=\langle x^2\rangle=\frac1{2a},

the score for the logarithmic covariance parameter is

SK(x)=∂∂log⁡Klog⁡Pa(x)=ax2−12.S_K(x)=\frac{\partial}{\partial\log K}\log P_a(x) =ax^2-\frac12.

It is centered and has positive Fisher norm:

⟨SK⟩=0,⟨SK2⟩=12.\langle S_K\rangle=0, \qquad \langle S_K^2\rangle=\frac12.

The subtraction −1/2-1/2 is forced by normalization. Thus vacuum or partition- function variation is not an untracked correction; it is already included in the score.

Consequence

The tangent of Entry 2211 has an actual boundary-field representative: a centered quadratic power insertion. It is not merely a derivative in an abstract coefficient space.

For independent momentum modes, the corresponding scores are orthogonal. Their Fisher metric is positive and diagonal before evaluation coincidences are imposed.

Evidence

View source on GitHub