Alphabeta Math
LemmaStatement: AI-adaptedProof: AI-adaptedPipeline-generatedprecheck passjudge 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.

Products of variables with Gaussian tails are sub-exponential

Statement

Let U,V be real random variables on one probability space, not necessarily independent, and suppose for some τ>0 that P(∣U∣>t)≤e−t2/(2τ),P(∣V∣>t)≤e−t2/(2τ)(t>0). Then W=UV−E[UV] is well-defined, centered, and sub-exponential with parameters (1024τ2,16τ) in the convention of Sub-Gaussian variables and vectors, and sub-exponential variables.

Facts & Assumptions

Given: Real random variables U,V with the displayed directional tail bounds and τ>0.

[F1]

Layer cake gives ET=∫0∞P(T>t) dt for nonnegative T (Layer-cake formulas for random variables).

[F2]

Sub-exponentiality with parameters (ν2,b) means EeλW≤eν2λ2/2 for ∣λ∣≤1/b (Sub-Gaussian variables and vectors, and sub-exponential variables).

[F3]

Increasing nonnegative measurable partial sums pass to their pointwise limit under expectation. (Monotone convergence for the integral)

[F4]

Measurable partial sums dominated in absolute value by one integrable random variable converge in L1 and in expectation to their pointwise limit. (Dominated convergence)

[F5]

Expectation is linear on finite sums of integrable random variables. (Linearity, monotonicity, and the modulus bound for expectation)

Proof

1.1F1givenalgebra

Applying [F1] to U2,V2 gives EU2,EV2≤2τ. Since 2∣UV∣≤U2+V2, UV is integrable and ∣EUV∣≤2τ. Also, for x>0, ∣UV∣>x implies ∣U∣>x or ∣V∣>x, so P(∣UV∣>x)≤2e−x/(2τ).

2.1step 1.1algebracases

Put W=UV−EUV. If t≥4τ, then ∣W∣>t implies ∣UV∣>t−∣EUV∣≥t/2, so step 1.1 gives P(∣W∣>t)≤2e−t/(4τ)≤4e−t/(8τ). If 0<t<4τ, then 4e−t/(8τ)>4e−1/2>1, so the same final bound follows from P(∣W∣>t)≤1. Thus for all t>0, P(∣W∣>t)≤4e−t/(8τ).

3.1F1step 2.1algebra

By [F1] and step 2.1, for every integer m≥2, E∣W∣m=m∫0∞tm−1P(∣W∣>t) dt≤4(8τ)mm!. The identity follows from substitution u=t/(8τ) and ∫0∞um−1e−udu=(m−1)!.

4.1F3F5step 1.1step 3.1algebra

Fix ∣λ∣≤1/(16τ) and put x=8τ∣λ∣≤1/2. The variable W is integrable and centered by step 1.1. The nonnegative measurable partial sums AN=∑m=0N∣λW∣m/m! increase pointwise to e∣λW∣. Finite linearity [F5] and monotone convergence [F3], applied only to these nonnegative partial sums, give Ee∣λW∣=1+∣λ∣E∣W∣+∑m≥2∣λ∣mE∣W∣mm!≤1+∣λ∣E∣W∣+4∑m≥2xm<∞. Thus e∣λW∣ is an integrable dominating variable.

5.1F2F4F5step 3.1step 4.1algebra∎

The signed partial sums SN=∑m=0N(λW)m/m! converge pointwise to eλW and satisfy ∣SN∣≤AN≤e∣λW∣. Dominated convergence [F4] and finite linearity [F5] justify taking expectations termwise; the scalar series of expectations is absolutely convergent by step 4.1. Since EW=0, step 3.1 gives EeλW=1+∑m≥2λmEWmm!≤1+4∑m≥2xm=1+4x21−x≤1+8x2≤e8x2=e512τ2λ2. By [F2] the parameters remain ν2=1024τ2 and b=16τ.

Depends on

Used by

Dependency tree · two levels

33 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