Proposition 5.4 — simultaneous concentration with probability
ProvedBilateralProfitMHR.proposition_5_4bilateral-trademonotone-hazard-rateonline-learningp2o-gran-per-chapterp2o-plan-paperp2o-strengthenedp2o-v1posted-pricesregret-bounds
Let , and run Initialize mhr (Algorithm 4) on fresh i.i.d. valuations from a product law whose densities are strictly positive on . With probability at least , simultaneously at every scale reached by the procedure, the estimates computed at that scale satisfy
Here is the triangle grid and the reached scales are those tested up to and including the trigger.
This clean event is the one on which the guarantees of Lemma 5.3 are proved.
Differences from the source:
- generalization: the MHR distributions of §5 → every independent pair with strictly positive densities on ; reason: the proof (pp.59–60) uses only independence and monotonicity of ; the statement specializes to the MHR class
- reading: "the first recursive split separating and " → the midpoint of the smallest bisection interval containing both, so ; reason: the bisection rule is not spelled out; this is the reading under which the proof holds as printed ("The construction includes , , and in ", p.60)
- encoding: posting each pair for fresh rounds → an array of independent samples indexed by (scale, grid pair, repetition); reason: the run at a scale reads only that scale's samples, so every event about the run has the same probability
- encoding: "with probability at least " → a measurable event of probability at least on which (29) holds at every reached scale; reason: same statement, stated without outer measure
Preamble
import Mathlib import Definitions.Def_BilateralProfitMHR_InitializeMHR
Formal statement
namespace BilateralProfitMHR
open MeasureTheory BilateralProfit MechanismDesign.BilateralTrade
open scoped ENNReal
/-- **Proposition 5.4 (Simultaneous concentration)**, Brüstle, di Gregorio, Fusco, Leonardi
(manuscript, sha256 e91e8127…), p.33 (PDF 35); proof App. A.5, pp.59–60 (PDF 61–62): let
`2 ≤ K ≤ T`. With probability at least `1 − 1/T`, simultaneously at every scale `δ` reached by
Initialize mhr, the estimates computed at that scale satisfy
`|µ̂(e) − µ(e)| ≤ (µ(e) + δ)/16` for every `e ∈ G_K` (29).
**Formalization Note.** `G_K` is Algorithm 4's triangle grid `Act K`; the reached scales are
`testedScales` (up to and including the trigger). The probability is over the fresh probing
valuations, i.i.d. with law `E.prior` (`initLaw`); "with probability at least `1 − 1/T`" is a
measurable event of that probability on which the bounds hold. Stated for every independent
pair of laws with positive densities on `[0, 1]`, MHR or not (the proof uses only independence
and monotonicity). -/
theorem proposition_5_4 (T K : ℕ) (hK : 2 ≤ K) (hKT : K ≤ T) (E : Environment)
(hE : UnitSupport E) :
∃ G : Set (Samples K T), MeasurableSet G ∧ 1 - (T : ℝ≥0∞)⁻¹ ≤ initLaw K T E.prior G ∧
∀ ω ∈ G, ∀ h ∈ testedScales K T ω, ∀ e : Act K,
|muHat K T h ω e - muP E.prior (pt K e.1)| ≤ (muP E.prior (pt K e.1) + delta h) / 16 := by sorry
end BilateralProfitMHR
Source
Brüstle, di Gregorio, Fusco, Leonardi, Regret Bounds for Profit Maximization in Bilateral Trade: Learning the Best Posted-Price Mechanism (manuscript, STOC 2027 submission, sha256 e91e81279505b765c7c96c219d674c0fbbbd6ca47a00da8fe228d498ca25ac87), Proposition 5.4, p.33 (PDF 35); proof App. A.5, pp.59–60 (PDF 61–62)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.