Alphabeta Math
TheoremStatement: Literature-sourcedProof: AI-adaptedPipeline-generatedjudge pass (gpt-6-luna)
How statement and proof provenance work

The first chip identifies the source of the statement or construction; the second identifies the source of its local proof or verification.

  • Literature-sourced: the exact statement appears in a cited source; only wording and notation differ.
  • AI-adapted: a semantically identical restatement of literature-sourced material, modulo indexing, notation, and boundary cases adopted by the library.
  • AI-generated: a genuinely novel statement formulated by AI, with no source for the claim itself.

These labels describe origin, not correctness: citations and verification chips remain separate evidence.

The support of a Banach Gaussian is the closure of its RKHS

Statement

Assume AC and countable choice. Let μ be a centered tight Borel Gaussian law on a real separable Banach space B, and let H⊆B be its covariance RKHS. Define supp⁡μ to be the set of x∈B for which every open neighborhood has positive μ-measure. Then supp⁡μ=H‾ B.

Facts & Assumptions

[F1]

Cameron–Martin translations by h∈H are equivalent to μ (Cameron–Martin translation formula for a separable Banach Gaussian).

[F2]

Tightness supplies a compact set of positive measure (Every borel probability on a polish space is tight).

[F3]

Centered multivariate normal vectors are determined by their covariance, and countable norming coordinates determine Borel laws on B (Multivariate normal law, including singular covariance, Characteristic function of a multivariate normal law, A countable norming family determines Borel laws).

[F4]

A point outside a closed linear subspace of a normed space can be separated from it by a continuous linear functional (Relative geometric Hahn–Banach with the exact open, closed, and compact hypotheses, The real dominated-extension principle as an additional hypothesis over ZF).

[A2]

Under AC, the real dominated-extension theorem supplies the Hahn–Banach principle HB used by geometric separation (Hahn-Banach dominated extension theorem for real vector spaces, The real dominated-extension principle as an additional hypothesis over ZF).

[F5]

Countable dense sets and rational-radius balls give countable bases for the separable Banach factors; their product Borel sigma-algebra is the product of the factor Borel sigma-algebras. Continuous maps have Borel preimages of Borel sets. (Separability: the existence of an at most countable dense subset, Borel sigma-algebra of a product of second-countable spaces, A continuous map has Borel preimages of Borel sets)

[F6]

A finite stack of independent multivariate Gaussian vectors is multivariate Gaussian, including singular covariance matrices. (A finite stack of independent Gaussian vectors is jointly Gaussian)

Proof

Given: AC, countable choice, a centered tight Borel Gaussian law μ on a real separable Banach space B and its covariance RKHS H.

1.1A1A2F2F3F5F6givenchoose

First, 0∈supp⁡μ. Fix ε>0. By [F2] choose compact K with μ(K)>1/2 and cover it by finitely many balls of radius ρ=ε/(22). One ball B(x,ρ) has positive μ-measure. On (B×B,B(B)⊗B(B),μ⊗μ) let W,W′ be the independent coordinate copies. By [F5], this product sigma-algebra is the Borel sigma-algebra of B×B; the maps (x,y)↦x−y and x↦2x are continuous, so W−W′ and 2W are Borel random elements. The event that both copies lie in B(x,ρ) has positive probability and implies ∥W−W′∥<2ρ. Choose a countable norming family using countable choice and HB from [A2]. For any finite subfamily, the coordinate vectors of W,W′ are independent centered multivariate normals with common covariance Σ; [F6] makes their stack multivariate Gaussian, so its linear image giving the difference is multivariate Gaussian. The vector for 2W is also multivariate Gaussian, and both vectors have covariance 2Σ. Thus [F3] makes their finite laws equal. Countable norming-law determination in [F3] gives W−W′=d2W. Hence μ(B(0,ε/2))>0, and therefore μ(B(0,ε))>0.

1.2A2F4givenalgebra

Suppose x∉H‾ B. By [F4] choose ℓ∈B∗ with ℓ(H)=0 and ℓ(x)≠0. For every m∈B∗, ℓ(Cm)=E[ℓ(W)m(W)]=0, since Cm∈H. Taking m=ℓ gives E[ℓ(W)2]=0, so ℓ(W)=0 almost surely. Hence μ(ker⁡ℓ)=1. Because ker⁡ℓ is closed and does not contain x, some open neighborhood of x is disjoint from ker⁡ℓ and has measure zero. Thus x∉supp⁡μ.

2.1F1step 1.1

For h∈H, [F1] gives equivalence of the law of W−h with μ. Thus its mass on B(0,ε) is positive by step 1.1. That mass is μ(B(h,ε)), so every h∈H lies in the support. Since the support is closed, H‾ B⊆supp⁡μ.

3.1step 2.1step 1.2∎

Step 2.1 gives H‾ B⊆supp⁡μ, and step 1.2 gives the reverse inclusion. Therefore the two sets are equal.

Depends on

Used by

Dependency tree · two levels

90 results within two dependency steps of this one, each drawn at its shortest distance from it. An arrow runs from a result to what uses it, so the chart reads left to right and ends at this result, which carries a heavier outline. Every node is a link to that result. Click elsewhere on the chart to enlarge it.

Sources