Proof of Theorem 3, p. 24 — a net of probability measures on a compact Hausdorff space has a regular cluster point
ProvedHartSchmeidler.Compact.exists_regular_cluster_pointLet be a compact Hausdorff space with its Borel σ-algebra, let be a nonempty directed set, and let be a net of probability measures on . Then there is a regular probability measure on that is a cluster point of the net in the topology induced by the continuous functions : for every finite family , every and every there is with
In the proof of Theorem 3 this is applied with (compact Hausdorff by Tychonoff), the f-sets containing a fixed profile, ordered by inclusion, and the correlated equilibria of the games ; the paper obtains from the Banach–Alaoglu theorem in the unit ball of , the dual of .
Formalization Note The cluster-point property is written out with finitely many test functions, which is exactly cluster-point convergence for the weak-* topology induced by . Regularity is Mathlib's Measure.Regular (outer regular, and inner regular by compact sets on open sets).
import Mathlib
namespace HartSchmeidler.Compact
open MeasureTheory
/-- Proof of Theorem 3, p. 24 (Banach–Alaoglu step): on a compact Hausdorff space with its Borel
σ-algebra, every net of probability measures indexed by a nonempty directed set has a cluster
point, for the topology induced by the continuous functions, that is a regular probability
measure. -/
theorem exists_regular_cluster_point {X : Type*} [TopologicalSpace X] [CompactSpace X]
[T2Space X] [MeasurableSpace X] [BorelSpace X]
{D : Type*} [Preorder D] [IsDirected D (· ≤ ·)] [Nonempty D]
(q : D → Measure X) (hq : ∀ d, IsProbabilityMeasure (q d)) :
∃ p : Measure X, IsProbabilityMeasure p ∧ p.Regular ∧
∀ (fs : Finset C(X, ℝ)) (ε : ℝ), 0 < ε → ∀ d₀ : D, ∃ d : D, d₀ ≤ d ∧
∀ f ∈ fs, |∫ x, f x ∂p - ∫ x, f x ∂(q d)| < ε := by sorry
end HartSchmeidler.Compact
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.