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.
Changing a conditional kernel on a null fiber leaves information unchanged
Statement
Assume AC. Let Z be uniform on the Borel unit interval and let X=Y=0 almost surely, with finite value space . The regular conditional kernels may be replaced by , for , and . The pointwise conditional KL between the conditional joint law of and the product of these marginal kernels changes from zero everywhere to at , but its -integral and remain zero.
Facts & Assumptions
Given: AC (and hence countable choice), the uniform Borel probability on , and almost surely.
Countable choice is required by the interval-volume formula used for and its null singleton; AC is inherited by the KL definitions. The Axiom of Countable Choice () The Axiom of Choice
and finite discrete spaces are standard Borel; the product of the two finite target spaces is also standard Borel. The unit interval with its Borel structure is standard Borel Standard Borel spaces Polish spaces are separable completely metrizable spaces Separability: the existence of an at most countable dense subset Finite products of standard Borel spaces are standard Borel
Under countable choice, Lebesgue measure is countably additive, so its Borel restriction is a measure. Borel subsets are Lebesgue measurable; the restriction to has mass one, the singleton has uniform measure zero, and its complement has measure one. Lebesgue measurable sets, the family , and the restricted set function Assuming countable choice, is a sigma-algebra containing every elementary set and is a complete measure extending elementary volume Assuming countable choice, every Borel subset of is Lebesgue measurable A box in with parameters is Lebesgue measurable of measure , whichever of its faces are included
Supplied regular conditional distributions that agree outside a conditioning-null set give the same same-marginal conditional-product law; conditional mutual information is its KL from the joint law, and KL of equal probability laws is zero by the unit density. The same-marginal conditional-product law is well-defined Conditional mutual information by a same-marginal product law Kullback–Leibler divergence
Conditional KL is the integral of the measurable pointwise kernel divergence. Conditional KL is measurable for supplied kernels Conditional KL for supplied probability kernels
For finite laws, KL is the atomwise sum with divergence when the first law charges an atom the reference law misses. Finite KL divergence agrees with measurable-space KL Finite distributions, entropy, divergence, and stochastic matrices Information units and logarithm-base convention
Proof
Given: The hypotheses in the Statement.
By [F2], the restriction is a probability and . On each finite target set, define distance 0 on equal points and 1 on distinct points; distinct endpoints force at least one intermediate distance to be 1, so the triangle inequality holds. Tolerance makes every Cauchy sequence eventually constant; the whole finite set is dense, and radius- balls are singletons, so the space is complete and discrete, hence Polish and standard Borel by [F1]. The product target is standard Borel by the finite-product supplier in [F1]. For any Borel and , replacing by changes only at , a null point. Thus both K and K' satisfy the conditional-law integral identity for X given Z; the constant kernel L satisfies it for Y, and satisfies it for . These kernels are measurable because their only nonconstant set of values is the Borel singleton .
Let be a conditional law of given Z. With the original kernels, for every z, so the pointwise KL is zero. With the alternative kernels, the product is at z=0 and for z>0. By [F5], because the first law charges and the reference gives it mass zero; for z>0 the divergence is zero. Thus the pointwise integrands differ only at a null point, and both conditional-KL integrals in [F4] are zero.
The joint law and the conditional-product reference law built from K,L agree. K' equals K -almost everywhere, so [F3] says the reference law built from K',L' is the same law. Consequently both versions give .
Depends on
- Kullback–Leibler divergence
- Conditional mutual information by a same-marginal product law
- Conditional KL for supplied probability kernels
- Conditional KL is measurable for supplied kernels
- The same-marginal conditional-product law is well-defined
- Finite KL divergence agrees with measurable-space KL
- Information units and logarithm-base convention
- Finite distributions, entropy, divergence, and stochastic matrices
- The unit interval with its Borel structure is standard Borel
- Lebesgue measurable sets, the family $\mathcal{L}(\mathbb{R}^n)$, and the restricted set function $\lambda_n$
- Assuming countable choice, $\mathcal{L}(\mathbb{R}^n)$ is a sigma-algebra containing every elementary set and $\lambda_n$ is a complete measure extending elementary volume
- A box in $\mathbb{R}^n$ with parameters $a_i\le b_i$ is Lebesgue measurable of measure $\prod_{i<n}(b_i-a_i)$, whichever of its faces are included
- Assuming countable choice, every Borel subset of $\mathbb{R}^n$ is Lebesgue measurable
- The Axiom of Countable Choice ($\mathrm{AC}_\omega$)
- The Axiom of Choice
- Standard Borel spaces
- Polish spaces are separable completely metrizable spaces
- Separability: the existence of an at most countable dense subset
- Finite products of standard Borel spaces are standard Borel
Used by
Nothing in the library uses this result yet.
Dependency tree · two levels
112 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
- Robert M. Gray, Entropy and Information Theory, corrected author-hosted first edition, Chapter 5 §§5.3 and 5.5 (standard reference, not scraped)
- Yury Polyanskiy and Yihong Wu, MIT 6.441 Information Theory, Chapter 2 §§2.4 and 2.6 (standard reference, not scraped)