Alphabeta Math
LemmaStatement: Literature-sourcedProof: AI-adaptedPipeline-generated
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.

KL is invariant under a measurable bijection

Statement

Let ϕ:(E,E)→(F,F) be a measurable bijection with measurable inverse. For probability measures P,Q on E, let ϕ#P,ϕ#Q be their push-forwards. Then D(ϕ#P∥ϕ#Q)=D(P∥Q).

Facts & Assumptions

Given: AC, the measurable isomorphism ϕ, and probability measures P,Q on E.

[A1]

AC supplies the Radon–Nikodym density in the absolutely continuous branch of KL. The Axiom of Choice

[F1]

KL is +∞ outside absolute continuity; under absolute continuity it is the integral of the log Radon–Nikodym density. Kullback–Leibler divergence

[F2]

Push-forward integration satisfies ∫g d(ϕ#P)=∫g∘ϕ dP for every nonnegative measurable g, and for integrable signed g. Change of variables for expectation

Proof

Given: The hypotheses in the Statement.

1.1F1givenchoose

If P≪̸Q, choose E_0 with Q(E0)=0<P(E0). Since ϕ−1 is measurable, ϕ(E0)=(ϕ−1)−1(E0) is measurable; then ϕ#Q(ϕ(E0))=0<ϕ#P(ϕ(E0)). Both divergences are +∞ by [F1].

1.2A1F1F2givenalgebra∎

If P≪Q, let r=dP/dQ. [given] The function s=r∘ϕ−1 is measurable. For every B in F, ∫Bs d(ϕ#Q)=∫ϕ−1(B)r dQ=P(ϕ−1(B))=ϕ#P(B), so s is a Radon–Nikodym density of ϕ#P with respect to ϕ#Q. By [F2], the push-forward integral identity applies separately to the nonnegative functions (ln⁡s)+ and (ln⁡s)−; the negative-part integrals are finite by [F1]. Therefore D(ϕ#P∥ϕ#Q)=∫(ln⁡s)+d(ϕ#P)−∫(ln⁡s)−d(ϕ#P)=∫(ln⁡r)+dP−∫(ln⁡r)−dP=D(P∥Q), including the case where the positive-part integral is +∞.

Depends on

Used by

Dependency tree · two levels

36 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