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.
Partial feature flows have exact energy and weight-moment identities
Statement
Assume AC. Let be either all of or the compact closure of a nonempty bounded convex Euclidean domain and let and be bounded maps, with separable Hilbert. Put , , and assume the feature-energy hypotheses and the prescribed differentiable risk . Let be any flow in the feature-energy Wasserstein definition. Write , , , and . On every finite interval these quantities have continuous or absolutely continuous representatives as appropriate, and These identities hold at every finite pair of endpoints. In particular and . If , then, with , No bounded-support assumption on the initial weights is required.
Statement references
AC implies DC implies countable choice
Feature-measure energy and its regularity assumptions
Wasserstein gradient flow for a feature-measure energy
Superposition for finite-action Euclidean continuity equations
Finite partition energy yields Hilbert-valued AC2 paths
Bounded linear maps commute with Bochner integration
Fundamental theorem of calculus for absolutely continuous functions
Facts & Assumptions
Given: AC and the hypotheses in the Statement.
A finite-action continuity-equation solution has a law of absolutely continuous characteristic paths with its time marginals and velocity. Superposition for finite-action Euclidean continuity equations
The prescribed velocity is the negative minimum of the full constrained local subdifferential. Wasserstein gradient flow for a feature-measure energy Feature-measure energy and its regularity assumptions
Hilbert paths with finite partition energy have Bochner derivatives, and bounded linear maps commute with their integrals. Scalar AC functions obey the endpoint FTC under DC and countable choice, both supplied by AC through the choice-implication theorem. Finite partition energy yields Hilbert-valued AC2 paths Bounded linear maps commute with Bochner integration Fundamental theorem of calculus for absolutely continuous functions The Axiom of Choice AC implies DC implies countable choice
Cauchy–Schwarz and the inherited AC assumption justify the path/time integrals. Cauchy–Schwarz: , with equality exactly for dependent pairs The Axiom of Choice
Proof
Given: The data and assumptions in the Statement.
The first derivatives of are bounded. This follows from compactness for bounded ; on the full Euclidean domain, boundedness and Lipschitz differential imply for unit , by Taylor's inequality on a unit segment. The scalar bound is identical. On a finite horizon, W2 continuity bounds the second moments of . Coupling Cauchy–Schwarz and show norm continuity of ; hence is bounded and continuous there.
Apply [F1]. Along almost every path, the derivative of is , with norm at most . Its path/time integral is finite by the uniform second-moment bound and Cauchy–Schwarz. The same bound controls the derivative of . For a real AC path, its derivative is zero almost everywhere on its zero set: at a differentiability point with nonzero derivative that zero is isolated, and isolated zeroes form a countable set. Thus the formula with sign zero set to zero holds almost everywhere. Integrating the pathwise FTC and using Fubini gives AC representatives of with the integrated derivatives. Their time derivatives are bounded by , hence are in on finite intervals; [F3] identifies the norm derivative of .
Compose the AC moment with : its differential is bounded on the compact image of the continuous moment path on the finite interval, so the composition is AC, and the ordinary Fréchet difference quotient gives derivative almost everywhere. Pair the inclusion [F2] with the characteristic velocity. At the potential derivatives are and . Any domain normal satisfies at two-sided differentiability times: test its normal inequality against for both signs of . At the angular velocity is zero, and step 2.1 gives almost everywhere on that zero set. The cusp and normal contributions therefore have zero pairing. The chain rule from step 2.1 yields almost everywhere. The established AC representative gives the first identity at all endpoints. Nonnegativity of gives and .
Off , the unconstrained weight equation is ; the theta-domain normal has no weight component. Consequently , also at almost every zero-set time. The product is integrable on finite intervals by Cauchy–Schwarz, the second-moment bound and bounded forcing. Integrating the pathwise identity gives the exact formula for . If moments are uniformly bounded by , the risk-sublevel bound gives , and . Thus almost everywhere; integrating proves the displayed linear growth bound.
Source note
The direct partial-model chain rule preserves the stationarity/escape claims for any flow in the definition. It does not apply the common-cutoff global theorem to an initial law outside its hypotheses.
Depends on
- The Axiom of Choice
- AC implies DC implies countable choice
- Feature-measure energy and its regularity assumptions
- Wasserstein gradient flow for a feature-measure energy
- Superposition for finite-action Euclidean continuity equations
- Finite partition energy yields Hilbert-valued AC2 paths
- Bounded linear maps commute with Bochner integration
- Fundamental theorem of calculus for absolutely continuous functions
- Cauchy–Schwarz: $|\langle x,y\rangle|\le\|x\|\,\|y\|$, with equality exactly for dependent pairs
Used by
Dependency tree · two levels
71 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.