OctoInformBeta

OpenAI preprint, September 2026 · §6. Equality of the two infima

Lemma 6.2 [Borel transport between absolutely continuous probabilities]: On standard Borel spaces, together with a uniform random variable on the source

Measurable transport from an atomless probability.

Hover over a blue or green term in the text, or a node in the graph, to find it in the other. Select it to read its declaration both in Lean and in words.

The Setting

Let XX and YY be standard Borel spaces, with YY nonempty, and let μ\mu and ν\nu be Borel probability measures on XX and YY, respectively. A probability μ\mu on XX is atomless when μ({x})=0\mu(\{x\})=0 for every x∈Xx\in X.

The pushforward Q#μQ_\#\mu of μ\mu by a measurable map Q:X→YQ:X\to Y is the probability measure determined by (1), for every Borel set A⊆YA\subseteq Y: (Q#μ)(A)=μ(Q−1(A)).(Q_\#\mu)(A)=\mu\bigl(Q^{-1}(A)\bigr).(1) A measurable transport Q:X→YQ:X\to Y from μ\mu to ν\nu is a measurable map whose pushforward is ν\nu.

Write I=[0,1]I=[0,1] and let λ\lambda be Lebesgue probability measure on II.

The Claim

Theorem 1. Let μ\mu be an atomless Borel probability measure on a standard Borel space XX, and let ν\nu be a Borel probability measure on a nonempty standard Borel space YY. There is a measurable map Q:X→YQ:X\to Y such that Q#μ=ν.Q_\#\mu=\nu.(2)✦ Let μ\mu be an atomless Borel probability measure on a standard Borel space XX. There is a measurable map f:X→If:X\to I such that f#μ=λ.f_\#\mu=\lambda.(3)✦

In the paper, the result reads: for an integer d≥2d\geq2 and absolutely continuous Borel probability measures μ\mu and ν\nu on ℝd\mathbb R^d, there is a Borel map Q:ℝd→ℝdQ:\mathbb R^d\to\mathbb R^d with Q#μ=νQ_\#\mu=\nu.

What Is Checked

The transport assertion (2) is stated over probability measures on standard Borel spaces with an atomless source and a nonempty target, and the uniform random variable assertion (3) uses an atomless standard Borel source and Lebesgue probability on II.