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.
The observed density is the latent marginal under Tonelli
Statement
Assume the hypotheses of Latent-variable joint and observed-data factorization. Then is jointly measurable in . For each fixed , it is an observed-data probability density: it is -measurable and Consequently for -almost every . On every valid fibre , is a measurable -density of total mass one. Finally, if is product measurable and for a fixed , Fubini permits exchanging the latent and observed integrals of . No exchange is claimed without this absolute integrability.
Facts & Assumptions
Given: The parameter and data spaces, sigma-finite measures , jointly measurable normalized , and, for the last clause, a product measurable complex-valued with .
Latent-variable joint and observed-data factorization: is jointly measurable, each fixed- joint density integrates to one, , and is defined only when .
Measure kernel and probability kernel: a uniformly sigma-finite kernel comes with a specified exhaustion for which each kernel mass on is finite.
Measurability of integration against a kernel: integrating a nonnegative product-measurable function against a finite or uniformly sigma-finite kernel gives a measurable function of the parameter.
Tonelli's theorem for nonnegative measurable functions on a sigma-finite product: for nonnegative product-measurable , its section integrals are measurable and the iterated and product integrals agree.
Fubini's theorem for L^1 functions on a sigma-finite product: for an function, almost every section is integrable and the iterated integrals equal the product integral.
Monotonicity and nonnegative homogeneity of the nonnegative integral: pointwise order and multiplication by a positive finite scalar pass to nonnegative integrals.
Every complete ordered field is Archimedean: the positive integers are unbounded in .
The Borel sigma-algebra on the extended real line: the extended-real Borel structure contains and its Borel subsets.
Threshold characterisations of real-valued and extended-real-valued measurability: an extended-real function is measurable if all strict upper threshold preimages are measurable.
Composition with a Borel measurable outer map preserves measurability: composition of a measurable map with a Borel measurable outer map is measurable.
Arithmetic and lattice operations preserve measurability whenever they are defined: products of extended-real measurable functions are measurable under its stated zero-times-infinity convention.
The product sigma-algebra and its finite iterates: the finite product sigma-algebra is generated by measurable rectangles.
Finite, sigma-finite, and semifinite measures: sigma-finiteness provides a countable measurable exhaustion by sets of finite measure.
Proof
Fix and let be the collection of subsets whose section belongs to . Sections commute with complements and countable unions, so is a sigma-algebra. It contains each measurable rectangle , whose section is if and empty otherwise. By [F12], every product-measurable set has a measurable section. Applying this to preimages of Borel sets shows that is product-measurable. For complex , first compose with the continuous Borel coordinate maps ; the same section argument makes both and measurable, hence is complex-measurable.
By [F13], choose a measurable exhaustion with . On , the constant assignment is a measure kernel: each section is and each evaluation map is constant. The same exhaustion makes it uniformly sigma-finite by [F2].
For fixed , step 1.1 makes product-measurable, so Tonelli [F4] gives by [F1]. Thus is a probability density. Also is measurable. For every positive integer , , so [F6] gives . Since this holds for every , [F7] implies .
The function is nonnegative and product-measurable by [F1]. Under the finite-product identification from [F12], apply [F3] to the constant kernel from step 1.2. It follows that is measurable on .
For fixed , write and . Define the finite-valued measurable function and then By [F11] this is complex-valued product-measurable; it is defined to be zero on the null set where . It agrees with off that null set, and the assumed absolute integrability gives . Fubini [F5] applies to this representative and gives with its zero assignment on exceptional null sections. Since almost everywhere on the fixed- product space, these are the asserted exchanges for .
Let , measurable by step 2.2. For each positive integer , , so [F6] and step 2.1 give . Since this holds for every , [F7] forces . If , division by that scalar and the definition of give . The valid-fibre set is measurable by step 2.2 and [F8]. For measurability, define by for finite and . For , ; for it is ; and for it is . These sets are Borel by [F8], so [F9] makes Borel. By [F10], is measurable on . Pulling it back along the measurable coordinate projection from the finite product space, which is measurable by [F12], gives a product-measurable function; multiplying by preserves measurability by [F11]. This product restricts to on , so is measurable on the valid-fibre set.
Steps 2.2, 2.1, 3.2, and 3.1 prove the marginal measurability, normalization, valid-fibre density, and integrable-exchange clauses.
Depends on
- The Borel sigma-algebra on the extended real line
- Finite, sigma-finite, and semifinite measures
- Latent-variable joint and observed-data factorization
- Measurable spaces and measurable sets
- Measure kernel and probability kernel
- The product sigma-algebra and its finite iterates
- Monotonicity and nonnegative homogeneity of the nonnegative integral
- Arithmetic and lattice operations preserve measurability whenever they are defined
- Composition with a Borel measurable outer map preserves measurability
- Fubini's theorem for L^1 functions on a sigma-finite product
- Measurability of integration against a kernel
- Every complete ordered field is Archimedean
- Threshold characterisations of real-valued and extended-real-valued measurability
- Tonelli's theorem for nonnegative measurable functions on a sigma-finite product
Used by
- Expectation-maximization Algorithm
- A finite mixture latent factorization Example
- A generalized EM step can increase Q without maximizing it Example
- A support-losing candidate has infinite directional KL Example
- A zero-normalizer latent fiber Example
- An EM likelihood with two strict local maxima Example
- Gaussian mixture likelihood is unbounded under covariance collapse Example
- Responsibilities in a two-component mixture Example
- EM does not decrease the observed log likelihood Theorem
Cited to discharge well-definedness by Latent-variable joint and observed-data factorization.
Dependency tree · two levels
52 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
- Mario A. T. Figueiredo, Lecture Notes on the EM Algorithm (2004) (standard reference, not scraped)