Documentation

LeanPool.AsymptoticTrianglePacking.Internal.Covered

LeanPool.AsymptoticTrianglePacking.Internal — Module C4b-3b (part 1) : probability a vertex is #

covered

Standalone, Mathlib-only. Foundation for the Rödl-nibble project.

Under a Bernoulli retention, a vertex x is covered in a round when it lies in some edge of the round's matching. Since a matched edge is in particular retained, the covered event is contained in the union of the retention events of the edges through x; a union bound gives

ℙ(x covered) ≤ deg(x) · p.

This union-bound estimate deliberately sidesteps the delicate correlation structure of the matching (we only use "matched ⟹ retained"), and feeds the residual-degree mean lower bound E[deg_residual(v)] ≥ deg(v)·(1 - r·d·p) (module C4b-3b part 2).

retainedSet ρ ω is the (classically decidable) set of retained edges at outcome ω. Must be placeholder-free and axiom-clean [propext, Classical.choice, Quot.sound].

noncomputable def LeanPool.AsymptoticTrianglePacking.Internal.retainedSet {V : Type u_1} [DecidableEq V] {Ω : Type u_2} [MeasureTheory.MeasureSpace Ω] (H : Finset (Finset V)) {p : ℝ} (ρ : BernoulliRetention H p) (ω : Ω) :

The set of retained edges at outcome ω (classically decidable membership in the events).

Equations
Instances For

    C4b-3b(1) — vertex-covered probability bound. The probability that a vertex x is covered by the round's matching is at most deg(x) · p.

    C4b-3b(2) — edge-hit probability. The probability that an edge e loses a vertex to the covered set (so it fails to survive into the residual) is at most ∑_{x∈e} deg(x)·p. Union bound over the vertices of e of prob_vertex_covered.

    theorem LeanPool.AsymptoticTrianglePacking.Internal.prob_edge_hit_le {V : Type u_1} [DecidableEq V] {Ω : Type u_2} [MeasureTheory.MeasureSpace Ω] {H : Finset (Finset V)} {p : ℝ} {r Δ : ℕ} (ρ : BernoulliRetention H p) (hr : Hypergraph.IsUniform H r) (hΔ : ∀ (x : V), Hypergraph.degree H x ≤ Δ) {e : Finset V} (he : e ∈ H) :

    C4b-3b(2') — edge-hit probability, regular form. For an r-uniform hypergraph with max degree ≤ Δ, an edge is hit with probability at most r·Δ·p.