Alphabeta Math
LemmaStatement: Literature-sourcedProof: AI-generatedPipeline-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.

Column-overlap formula for the zero-sum ℓ1 coefficient

Statement

For a finite column-stochastic matrix P, define δ(P)=sup⁡{∥Pz∥1:z∈H, ∥z∥1=1} when n≥2, and set δ(P)=0 when n=1 (then H={0}). If pj denotes column j, then

δ(P)=12max⁡j,k∥pj−pk∥1=1−min⁡j,k∑imin⁡(Pij,Pik).

Thus δ(P) is a computable induced contraction factor on the zero-sum subspace H; for every z∈H, ∥Pz∥1≤δ(P)∥z∥1.

Facts & Assumptions

Given: A finite column-stochastic matrix P and the zero-sum subspace H from Finite column-stochastic matrices and probability vectors.

[F1]

Finite column-stochastic matrices and probability vectors: every column pj is nonnegative with coordinate sum one, H={z:∑izi=0}, and ∥z∥1=∑i∣zi∣.

[F2]

Column-stochastic matrices preserve mass and contract ℓ1: P preserves coordinate sums and satisfies ∥Pz∥1≤∥z∥1 for every real vector z.

Proof

Given: The finite matrix P and its columns p1,…,pn.

Proof technique: decompose a zero-sum vector into positive and negative probability masses.

1.1F1givenalgebra

Let z∈H with ∥z∥1=1. Write z=z+−z− coordinatewise, where zi+=max⁡(zi,0) and zi−=max⁡(−zi,0). Since ∑izi=0, the positive and negative masses agree, and their sum is one; hence ∑izi+=∑izi−=1/2. Therefore u=2z+ and v=2z− are probability vectors and Pz=(Pu−Pv)/2.

2.1step 1.1F1algebra

We have Pu=∑jujpj and Pv=∑kvkpk. Thus Pz=12∑j,kujvk(pj−pk). The coefficients ujvk are nonnegative and sum to one, so the triangle inequality gives ∥Pz∥1≤12max⁡j,k∥pj−pk∥1. Taking the supremum over such z proves the upper bound.

3.1F1step 2.1choosealgebra

If n≥2 and the maximum is positive, a maximizing pair has j≠k; if the maximum is zero, any distinct pair maximizes it. Choose such a distinct pair and set z=(ej−ek)/2. Then z∈H, ∥z∥1=1, and Pz=(pj−pk)/2, so the supremum is at least 12max⁡j,k∥pj−pk∥1. If n=1, both sides are zero by definition. This proves the first equality.

4.1F1step 3.1algebra

For two nonnegative probability columns a,b, ∣ai−bi∣=ai+bi−2min⁡(ai,bi). Summing gives 12∥a−b∥1=1−∑imin⁡(ai,bi). Apply this to every pair of columns; taking the maximum of the left side is one minus the minimum overlap on the right.

5.1F2step 3.1step 4.1givenalgebra∎

If n≥2, [F2] bounds every norm in the defining supremum by one, so 0≤δ(P)≤1; if n=1, δ(P)=0 by definition. For nonzero z∈H, apply the definition of δ(P) to z/∥z∥1 and use homogeneity; for z=0 the inequality is immediate. Thus ∥Pz∥1≤δ(P)∥z∥1 on all of H.

Depends on

Used by

Dependency tree · two levels

3 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