Alphabeta Math
LemmaStatement: AI-adaptedProof: 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.

ULA invariant laws have uniform exponential moments on the needed step ranges

Statement

Assume AC and the hypotheses of ULA's Gaussian kernel satisfies an exponential drift bound. Let Rγ be the ULA kernel, V=eU/2, W=eU/4, and γ0=(2L)−1. For every 0<γ<L−1 there is a unique invariant probability πγ, and πγV<∞. There are finite constants MV,MW,Mg, independent of γ, such that sup⁡0<γ≤γ0πγV≤MV,sup⁡0<γ<L−1πγW≤MW,∥∇U(x)∥2+∥x−x⋆∥2≤MgV(x).

Facts & Assumptions

Given: AC, the stated potential and radial constants, and 0<γ<L−1.

[A1]

AC is inherited through the Gaussian and weighted Hahn/Jordan interfaces. The Axiom of Choice

[F1]

The ULA kernel has a positive continuous Gaussian density, with proposal x−γ∇U(x)+2γZ. Unadjusted Langevin algorithm (ULA) The ULA Gaussian proposal is a probability kernel with a positive density

[F2]

The fixed-step Gaussian kernel satisfies RγV≤e−γV+bγ1Dγ for a finite constant and a compact ball. ULA's Gaussian kernel satisfies an exponential drift bound

[F3]

Radial drift gives U(x)≥c∥x−x⋆∥α−C for some c>0; hence V-sublevel sets are contained in compact balls. Radial gradient drift forces potential coercivity

[F4]

Global geometric drift and minorization on a sufficiently large weight-sublevel set give invariant existence, a finite weight moment, and uniqueness among all invariant probabilities. Global drift and one-step minorization give weighted contraction and a unique invariant law

[F5]

An exponential with positive exponent dominates every fixed nonnegative integer power at infinity. The exponential dominates every fixed nonnegative integer power at +∞

Proof

Given: The hypotheses in the Statement.

1.1A1F1F2F3

Fix γ. Dropping the indicator in [F2] gives a global drift RγV≤e−γV+bγ with finite bγ. Each sufficiently large V-sublevel is contained in a compact ball by [F3]. On the product of that starting ball and any fixed positive-radius target ball, the continuous positive density [F1] has a positive minimum: explicitly its Gaussian exponent is bounded because both balls and the continuous gradient are bounded. Restriction to the target ball and normalization of its positive finite volume give a common minorization. These are precisely the drift and minorization hypotheses for the weight V.

1.2F6givenalgebra

By [F6], g(x⋆)=0, and therefore Lipschitz continuity gives ∥g(x)∥≤L∥x−x⋆∥. If r=∥x−x⋆∥ is positive and at least the radial radius, Cauchy–Schwarz and the radial hypothesis give ∥g(x)∥r≥⟨g(x),x−x⋆⟩≥ρrα, so ∥g(x)∥≥ρrα−1. For h=y−x, integrate the derivative along x+th and subtract ⟨g(x),h⟩ to obtain U(y)−U(x)−⟨g(x),h⟩=∫01⟨g(x+th)−g(x),h⟩dt≤L2∥h∥2.

1.3F7algebra

Completing the square, [F7] gives for c<1/2 and scalar v the identity (2π)−1/2∫Re−(1/2−c)z2+vzdz=(1−2c)−1/2ev2/(2(1−2c)). Finite multiplication and Tonelli therefore give Eec∥Z∥2+⟨v,Z⟩=(1−2c)−d/2e∥v∥2/(2(1−2c)) for Z∼Nd(0,Id).

2.1F4step 1.1

Apply [F4] to the fixed-step kernel and weight V verified in step 1.1. It supplies a unique invariant probability πγ with finite V moment.

2.2F1step 1.2step 1.3algebra

Put g=∇U(x) and Hs=esU for 0<s<1 with 2sLγ<1. Apply step 1.2 to Y=x−γg+2γZ, multiply by s, and exponentiate. Step 1.3, with quadratic coefficient sLγ and linear coefficient s2γ(1−Lγ)g, gives RγHs(x)Hs(x)≤(1−2sLγ)−d/2exp⁡ ⁣(−sγ(1−s−Lγ/2)1−2sLγ∥g∥2). The coefficient follows by combining sγ(−1+Lγ/2) with s2γ(1−Lγ)2/(1−2sLγ) over the common denominator. All Gaussian integrals are finite under 2sLγ<1.

3.1step 2.1step 1.2step 2.2algebra

For s=1/2, step 2.2 reduces to RγV≤Vexp⁡ ⁣(d2[−log⁡(1−Lγ)]−γ4∥g∥2). If γ≤γ0, then −log⁡(1−Lγ)≤2Lγ, since −log⁡(1−z)=∫0z(1−u)−1du≤2z for z≤1/2. Choose a radius R at least the radial radius and large enough that ρ2R2(α−1)/4≥dL+1. Outside this ball step 1.2 makes the exponent at most −γ. Inside it let M=sup⁡∥x−x⋆∥≤RV(x)<∞. Since eaγ−e−γ=∫−γaγeudu≤(a+1)γeaγ0 for a=dL, the preceding bound gives RγV≤e−γV+bγ,b=M(dL+1)edLγ0, uniformly for small steps. Integrate against the finite-V invariant law from step 2.1 to obtain πγV≤bγ/(1−e−γ)≤beγ0, using 1−e−γ=∫0γe−udu≥γe−γ0. This proves the first uniform bound.

4.1step 2.1step 1.2step 2.2step 3.1algebra

For s=1/4 and 0<γ<L−1, the logarithm of the prefactor in step 2.2 is at most dLγ/2, by the same integral estimate with z=Lγ/2<1/2. Moreover s(1−s−Lγ/2)/(1−2sLγ)≥1/16, because the numerator is at least 1/16 and the denominator is at most one. Choose R' at least the radial radius such that ρ2(R′)2(α−1)/16≥dL/2+1. Outside this ball RγW≤e−γW; inside use M′=sup⁡∥x−x⋆∥≤R′W(x) and the exponential-difference estimate from step 3.1 with a=dL/2 and upper horizon L−1. Thus RγW≤e−γW+b′γ,b′=M′(dL/2+1)ed/2, for the whole step range. Since W≤V, its invariant moment is finite by step 2.1. Integrating and using 1−e−γ≥γe−1/L gives πγW≤b′e1/L, uniformly for every allowed step.

5.1F3F5step 1.2step 3.1step 4.1algebra∎

Put r=∥x−x⋆∥. For r≥1, [F3] gives V(x)≥e−C/2ecrα/2≥e−C/2ecr/2 because α>1. By [F5], r2/ecr/2 is bounded on [1,∞); on [0,1] use r2≤1≤V. Thus r2≤AV for some finite A. The Lipschitz bound in step 1.2 gives ∥∇U(x)∥2≤L2r2. Taking Mg=(1+L2)A, MV=beγ0 and MW=b′e1/L proves all asserted estimates. All constants depend only on the fixed potential and its displayed parameters.

Depends on

Used by

Dependency tree · two levels

97 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