gpa_lhs_upper
lemmaverified
Upper bound on the simplex: u*(σ) - h(σ) ≤ 1 - 1/α + 1/k for every σ ∈ Δ_k, where α ≥ 1 bounds every independent set.
Statement
open Finset in lemma gpa_lhs_upper {k : ℕ} (G : SimpleGraph (Fin k)) [DecidableRel G.Adj] (a : ℕ) (ha : 1 ≤ a) (hα : ∀ S, gpa_indep G S → S.card ≤ a) (σ : Fin k → ℝ) (hσ : σ ∈ gpa_simplex k) : gpa_uStar (gpa_adjMat G) σ - gpa_cost σ ≤ 1 - 1 / (a : ℝ) + 1 / (k : ℝ)
- Source
- step towards gpa_eq54_false / gpa_eq54_fails_of_indep (Generalized Principal-Agency: Contracts, Information, Games and Beyond, WINE 2024 (arXiv:2209.01146v2), Sec. 5.3, Eq. (5.4))
- Verified
- 06 Oct 2026
- Axioms
Classical.choiceQuot.soundpropext- Built on 6
- gpa_adjMatThe 0/1 matrix e of arXiv:2209.01146, Section 5.3 (taken from the maximum-independent-set instances of the cited convex-maximisation…
- gpa_uStarThe principal's utility of arXiv:2209.01146, Section 5.3: u*(σ) = Σ_i max{σ_i - Σ_j e_{ij} σ_j, 0} for a 0/1 matrix e.
- gpa_costThe convex signal cost of arXiv:2209.01146, Section 5.3: h(σ) = max{‖σ‖_∞ - 1/k, 0} (the norm on Fin k → ℝ is the sup norm).
- gpa_simplexThe probability simplex Δ_k.
- gpa_indepAn independent set of the graph G.
- gpa_uStar_le_indepKey structural lemma: for a nonnegative vector, the coordinates with a positive term of u* form an independent set P, and u*(σ) ≤ Σ_{i ∈…
- Used by 1
- gpa_lhs_isGreatestExact value of the left side of Eq. (5.4) of arXiv:2209.01146 on the independent-set instances: max_{σ ∈ Δ_k} u*(σ) - h(σ) = 1 - 1/α(G) +…
Lean source view module on GitHub
Choose a tactic — a marked line — to see the goals Lean had before and after it, as in Lean’s infoview. Step with ← → or j k.
open Finset in/-- Upper bound on the simplex: `u*(σ) - h(σ) ≤ 1 - 1/α + 1/k` for every `σ ∈ Δ_k`, where `α ≥ 1` bounds every independent set. -/lemma gpa_lhs_upper {k : ℕ} (G : SimpleGraph (Fin k)) [DecidableRel G.Adj] (a : ℕ) (ha : 1 ≤ a) (hα : ∀ S, gpa_indep G S → S.card ≤ a) (σ : Fin k → ℝ) (hσ : σ ∈ gpa_simplex k) : gpa_uStar (gpa_adjMat G) σ - gpa_cost σ ≤ 1 - 1 / (a : ℝ) + 1 / (k : ℝ) := by obtain ⟨hnn, hsum⟩ := hσ obtain ⟨P, hP, hle⟩ := gpa_uStar_le_indep G σ hnn set M := ‖σ‖ with hM have hcoord : ∀ i, σ i ≤ M := fun i => by have := norm_le_pi_norm σ i rw [Real.norm_eq_abs, abs_of_nonneg (hnn i)] at this exact this -- Σ_P σ ≤ 1 have hP1 : ∑ i ∈ P, σ i ≤ 1 := by rw [← hsum] exact Finset.sum_le_sum_of_subset_of_nonneg (Finset.subset_univ P) (fun i _ _ => hnn i) -- Σ_P σ ≤ a M have hM0 : 0 ≤ M := norm_nonneg σ have hPa : ∑ i ∈ P, σ i ≤ (a : ℝ) * M := by have h1 : ∑ i ∈ P, σ i ≤ ∑ _i ∈ P, M := Finset.sum_le_sum (fun i _ => hcoord i) have h2 : (P.card : ℝ) ≤ a := by exact_mod_cast hα P hP simp at h1 nlinarith have hcost : M - 1 / (k : ℝ) ≤ gpa_cost σ := le_max_left _ _ have haR : (1 : ℝ) ≤ a := by exact_mod_cast ha have hapos : (0 : ℝ) < a := by linarith by_cases hcase : 1 ≤ (a : ℝ) * M · -- M ≥ 1/a have hMa : 1 / (a : ℝ) ≤ M := by rw [div_le_iff₀ hapos]; linarith linarith · replace hcase := not_le.mp hcase -- u* - h ≤ aM - M + 1/k = (a-1)M + 1/k ≤ (a-1)/a + 1/k have hkey : ((a : ℝ) - 1) * M ≤ 1 - 1 / (a : ℝ) := by have hMlt : M < 1 / (a : ℝ) := by rw [lt_div_iff₀ hapos]; linarith have : ((a : ℝ) - 1) * M ≤ ((a : ℝ) - 1) * (1 / a) := mul_le_mul_of_nonneg_left hMlt.le (by linarith) have e : ((a : ℝ) - 1) * (1 / a) = 1 - 1 / (a : ℝ) := by field_simp linarith nlinarith