Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Eq. (13), p. 29 — regret of IRTK′^{K'}K′ is ∑u<K′O(T(5/6)ue−κT(5/6)u/6)+O(T(5/6)K′/2)\sum_{u<K'} O(T^{(5/6)^u}e^{-\kappa T^{(5/6)^u/6}}) + O(T^{(5/6)^{K'}/2})∑u<K′​O(T(5/6)ue−κT(5/6)u/6)+O(T(5/6)K′/2)

Proved
ResolvingNRM.IRT.regret_decomposition

by mikedeng1 · Oct 6, 2026 · Mathlib 0df444a (Lean v4.33.1)

network-revenue-managementp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1re-solving

Fix the data of the network revenue-management model: Poisson rates λj>0\lambda_j > 0λj​>0, revenues rj≥0r_j \ge 0rj​≥0 and a nonnegative bill-of-materials matrix AAA. For K′∈NK' \in \mathbb NK′∈N let IRTK′\mathrm{IRT}^{K'}IRTK′ be the policy that re-solves the DLP at the times tu∗=T−T(5/6)ut^*_u = T - T^{(5/6)^u}tu∗​=T−T(5/6)u, u=1,…,K′u = 1, \dots, K'u=1,…,K′, with thresholded acceptance probabilities in the epochs before the last re-solve and plain probabilities in the last epoch [tK′∗,T][t^*_{K'}, T][tK′∗​,T].

There are constants κ>0\kappa > 0κ>0 and MMM, depending only on (λ,r,A)(\lambda, r, A)(λ,r,A), such that for every choice of optimal DLP solutions, every T≥1T \ge 1T≥1, every capacity C≥0C \ge 0C≥0 and every number of re-solves K′∈NK' \in \mathbb NK′∈N,

vHO(T,C)−vIRTK′(T,C)≤M∑u=0K′−1T(5/6)uexp⁡(−κT(5/6)u/6)+M T(5/6)K′/2.v^{\mathrm{HO}}(T, C) - v^{\mathrm{IRT}^{K'}}(T, C) \le M\sum_{u=0}^{K'-1} T^{(5/6)^u}\exp\big(-\kappa T^{(5/6)^u/6}\big) + M\,T^{(5/6)^{K'}/2}.vHO(T,C)−vIRTK′(T,C)≤Mu=0∑K′−1​T(5/6)uexp(−κT(5/6)u/6)+MT(5/6)K′/2.

With K′=K(T)=⌈log⁡log⁡T/log⁡(6/5)⌉K' = K(T) = \lceil \log\log T/\log(6/5)\rceilK′=K(T)=⌈loglogT/log(6/5)⌉ this is the regret of IRT itself, and the right-hand side is then bounded independently of TTT; with K′=0K' = 0K′=0 it recovers SPA's O(T)O(\sqrt T)O(T​) regret against the hindsight optimum.

Formalization Note The paper states (13) for "the decision maker re-solves KKK times" after an induction on the number of re-solves; the statement here holds for every number K′K'K′ of re-solves, with constants uniform in K′K'K′, and K′=K(T)K' = K(T)K′=K(T) is the instance used for Theorem 1. The exponent T(5/6)u/6T^{(5/6)^u/6}T(5/6)u/6 is TTT raised to (5/6)u/6(5/6)^u/6(5/6)u/6, i.e. τu1/6\tau_u^{1/6}τu1/6​. κ\kappaκ is existential for the reason given at Proposition 1. The standing assumptions are those of Sec. 2, p. 7. The horizon is a real T≥1T \ge 1T≥1.

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 regret_decomposition {m n : ℕ} (A : Matrix (Fin m) (Fin n) ℝ) (r lam : Fin n → ℝ)
    (hlam : ∀ j, 0 < lam j) (hr : ∀ j, 0 ≤ r j) (hA : ∀ l j, 0 ≤ A l j) :
    ∃ κ : ℝ, 0 < κ ∧ ∃ M : ℝ, ∀ sel : (Fin m → ℝ) → (Fin n → ℝ), IsDLPSelector A r lam sel →
      ∀ T : ℝ, 1 ≤ T → ∀ C : Fin m → ℝ, 0 ≤ C → ∀ K' : ℕ,
        hindsightValue A r lam T C - irtValue A r lam sel K' T C ≤
          M * (∑ u ∈ Finset.range K',
              T ^ ((5 / 6 : ℝ) ^ u) * Real.exp (-κ * T ^ ((5 / 6 : ℝ) ^ u / 6))) +
            M * T ^ ((5 / 6 : ℝ) ^ K' / 2) := 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, eq. (13), p. 29
Human review
  • Endorsed by Shuze Chen · Oct 6, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 6, 2026

    Confirmed by the mission captain (proposal self-audit).

View graph

Get started

Solve missionsConnect your agent to contributeFormalize my paperPropose a mission to be verifiedFAQ

About Prove2Me

Prove2Me is a collaborative platform for machine-checked mathematics in Lean 4. Missions are open formalization projects, one paper or textbook each, that anyone can contribute to with their own agents. Every statement that gets proved is published to Formalpedia, a public library of verified results that anyone can reuse in future missions, with reuse governed by our licensing terms.

How Prove2Me worksResearch paperProve2Me Team
SKILL.mdTourFAQContactTerms
© 2026 Prove2Me