Alphabeta Math
LemmaStatement: Literature-sourcedProof: AI-adaptedPipeline-generatedjudge pass (gpt-6-luna)
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.

Hellinger distance is a metric and controls L1 and total variation

Statement

For probability densities p,q relative to one fixed measure μ, define h as in Root-L2 Hellinger distance between density laws and set d1(P,Q):=∫∣p−q∣ dμ. Then h is a metric on the induced probability laws and h(P,Q)2≤d1(P,Q)=2∥P−Q∥TV,d1(P,Q)≤2h(P,Q), where library total variation is the event supremum ∥P−Q∥TV=sup⁡A∈A∣P(A)−Q(A)∣.

Facts & Assumptions

Given: A measure space (X,A,μ) and nonnegative measurable densities p,q with integrals one; let P,Q be their induced probability laws.

[F1]

Root-L2 Hellinger distance between density laws: h(P,Q)=∥p−q∥2 for this fixed dominating measure.

[F2]

Minkowski's inequality for integrals, including p=∞: for L2(μ) functions, ∥f+g∥2≤∥f∥2+∥g∥2.

[F3]

Holder's inequality for integrals, including the endpoint cases: ∫∣fg∣ dμ≤∥f∥2∥g∥2 when both functions are in L2(μ).

[F4]

Total variation distance for probability laws: ∥P−Q∥TV=sup⁡A∈A∣P(A)−Q(A)∣.

[F5]

A nonnegative measurable function has integral 0 exactly when it vanishes almost everywhere: a nonnegative measurable function has zero integral exactly when it vanishes μ-almost everywhere.

[F6]

The Lebesgue integral is linear on L1(μ): integration is linear for integrable functions; here p−q∈L1(μ) because ∣p−q∣≤p+q, both densities integrate to one, and [F9] plus [F10] bound the integral of p+q by 2.

[F7]

Arithmetic and lattice operations preserve measurability whenever they are defined: sums, differences, products, and absolute values of measurable functions are measurable.

[F8]

Composition with a Borel measurable outer map preserves measurability: Borel functions of measurable real-valued functions are measurable; in particular this applies to square root on [0,∞), the sign function, and threshold indicators.

[F9]

Monotonicity and nonnegative homogeneity of the nonnegative integral: the nonnegative integral preserves pointwise order.

[F10]

Additivity of the nonnegative Lebesgue integral: the nonnegative integral is additive on nonnegative measurable functions.

Proof

1.1F1F2F5F6F7F8F9F10givenalgebra

Since p,q are measurable, [F8] makes p,q measurable and [F7] makes their difference and absolute value measurable. Their unit integrals put both roots in L2(μ). The triangle inequality in [F2] gives symmetry, nonnegativity, and the triangle inequality for h. If h(P,Q)=0, [F5] gives p=q almost everywhere, hence p=q almost everywhere and P=Q. Conversely, put g=p−q∈L1(μ) by [F6]. If P=Q, then for every measurable A, ∫g1A dμ=∫p1A dμ−∫q1A dμ=P(A)−Q(A)=0 by [F6]. For each integer m≥1, the set Em:={g≥1/m} is measurable by [F8]. Define um:=g1Em; it is nonnegative and measurable by [F7], and linearity gives ∫um dμ=0. By [F5], um=0 almost everywhere; since um≥1/m on Em, μ(Em)=0. Applying the same argument to (q−p)1{q−p≥1/m} shows μ({g≤−1/m})=0. Since {g≠0} is the countable union of these null sets, p=q almost everywhere and h(P,Q)=0. Thus h is a metric on these laws.

1.2F4F7F8F9F10givenalgebra

Put g:=p−q and A:={x:g(x)≥0}, which is measurable by [F7, F8]. Since ∫g dμ=0, the positive and negative parts have equal integral, so ∫Ag dμ=12∫∣g∣ dμ. For every measurable B, positivity of g on A and nonpositivity on Ac, together with monotonicity [F9], give ∫Acg≤∫Bg≤∫Ag; equality in the absolute supremum is attained at A (or Ac). Thus [F4] gives ∥P−Q∥TV=12d1(P,Q).

1.3F1F7F8F9givenalgebra

The functions in the pointwise inequality are measurable by [F7, F8]. If p≥q, then (p−q)2≤p−q, because 0≤(p−q)/(p+q)≤1 when the denominator is positive; the case q≥p is symmetric, and when both vanish the inequality is zero on both sides. Monotonicity [F9] and [F1] give h(P,Q)2≤d1(P,Q).

2.1F1F3F7F8F9step 1.1step 1.2algebra∎

By [F7, F8], the displayed products and sums of square roots are measurable. Factor ∣p−q∣=∣p−q∣(p+q). By [F3], d1(P,Q)≤h(P,Q)∥p+q∥2. Also ∥p+q∥22=2+2∫pq dμ≤4, again by [F3] and the unit L2 norms. Hence d1(P,Q)≤2h(P,Q), and step 1.2 gives ∥P−Q∥TV≤h(P,Q).

Depends on

Used by

Dependency tree · two levels

39 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