p. 30 — with ,
ProvedResolvingNRM.IRT.tau_Kirt_le_exp_oneanalysisp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1re-solving
Let be the number of re-solves of the IRT policy. For every real ,
In words: the time remaining at the last re-solving time of IRT, , is at most , so the last epoch of IRT, where no thresholds are used, has bounded length.
Formalization Note For the printed formula for is undefined (, ) or negative; with Lean's conventions and for , there and the claim reads .
Preamble
import Mathlib import Definitions.Def_RLPBidPrice_Unbiased_Model import Definitions.Def_ResolvingNRM_IRT_Model open RLPBidPrice.Unbiased Matrix
Formal statement
namespace ResolvingNRM.IRT
theorem tau_Kirt_le_exp_one (T : ℝ) (hT : 1 ≤ T) :
T ^ ((5 / 6 : ℝ) ^ Kirt T) ≤ Real.exp 1 := by sorry
end ResolvingNRM.IRT
Source
Bumpensanti, Wang, A Re-solving Heuristic with Uniformly Bounded Loss for Network Revenue Management, arXiv:1802.06192v3, Appendix B.1, p. 30, first line ('using the fact that T^{(5/6)^K} ≤ e'); K defined in Algorithm 3, p. 15
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.