§4.1.1, p. 5555 — for t ≥ 5, each experimental price is charged at least ¼√t times in periods 1, …, t
ProvedBanKeskin.UnknownSparsity.experiment_countdynamic-pricingp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1price-experimentation
Let and be the experimentation periods of the schedule (7) of Ban and Keskin (2021), in which the experimental prices and are charged. Then for every ,
This guarantees that the information about the price sensitivity grows at rate , which underlies the estimation-error bound of Lemma 3 and hence Theorem 3. ILQX uses the same schedule as ILSX (§4.1.1).
Formalization Note. Periods are 1-based. "Each experimental price is charged" counts the experimentation periods of that price, .
Preamble
import Mathlib import Definitions.Def_BanKeskin_UnknownSparsity_Model
Formal statement
namespace BanKeskin.UnknownSparsity
open Classical in
theorem experiment_count (t : ℕ) (ht : 5 ≤ t) :
(1 / 4 : ℝ) * Real.sqrt t ≤ (((Finset.Icc 1 t).filter fun k => BanKeskin.KnownSparsity.inM1 k).card : ℝ) ∧
(1 / 4 : ℝ) * Real.sqrt t ≤ (((Finset.Icc 1 t).filter fun k => BanKeskin.KnownSparsity.inM2 k).card : ℝ) := by sorry
end BanKeskin.UnknownSparsity
Source
Ban and Keskin, Personalized Dynamic Pricing with Machine Learning, Management Science 67(9) (2021), p. 5555, §4.1.1, after (7)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.