Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 7.1.2 — P(L_N ≥ M+u) ≤ 2exp(−u²/(4(M+u))) and P(L_N ≤ M−u) ≤ 2exp(−u²/(4M))

Proved
TalagrandConc.Subsequences.lis_concentration

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

concentration-of-measurelongest-increasing-subsequencep2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-bookp2o-v1probability

Let X1,…,XNX_1,\dots,X_NX1​,…,XN​ be independent random variables on [0,1][0,1][0,1], uniformly distributed or, more generally, distributed according to a common non-atomic probability μ\muμ. Let LN=LN(X1,…,XN)L_N = L_N(X_1,\dots,X_N)LN​=LN​(X1​,…,XN​) be the length of the longest increasing subsequence of X1,…,XNX_1,\dots,X_NX1​,…,XN​ (the largest ppp with i1<⋯<ipi_1 < \dots < i_pi1​<⋯<ip​ and Xi1≤⋯≤XipX_{i_1} \le \dots \le X_{i_p}Xi1​​≤⋯≤Xip​​), and let M=MNM = M_NM=MN​ be a median of LNL_NLN​. Then for all u>0u > 0u>0,

P(LN≥M+u)≤2exp⁡(−u24(M+u)),(7.1.3)P(L_N \ge M + u) \le 2\exp\Big(-\frac{u^2}{4(M+u)}\Big), \tag{7.1.3}P(LN​≥M+u)≤2exp(−4(M+u)u2​),(7.1.3) P(LN≤M−u)≤2exp⁡(−u24M).(7.1.4)P(L_N \le M - u) \le 2\exp\Big(-\frac{u^2}{4M}\Big). \tag{7.1.4}P(LN​≤M−u)≤2exp(−4Mu2​).(7.1.4)

Since LNL_NLN​ is distributed like the longest increasing subsequence of a uniform random permutation and MNM_NMN​ grows like 2N2\sqrt N2N​, these bounds show that LNL_NLN​ fluctuates around its median on the scale N1/4N^{1/4}N1/4, sharper than the scale N\sqrt NN​ given by martingale methods.

Formalization Note The variables are the coordinates of [0,1]N[0,1]^N[0,1]N under P=μ⊗NP = \mu^{\otimes N}P=μ⊗N, where μ\muμ is any probability measure on [0,1][0,1][0,1] with μ({t})=0\mu(\{t\}) = 0μ({t})=0 for every ttt (this includes Lebesgue measure). When M=0M = 0M=0 (only possible for N=0N = 0N=0) Lean's division gives u2/(4M)=0u^2/(4M) = 0u2/(4M)=0, so (7.1.4) reads P(LN≤−u)≤2P(L_N \le -u) \le 2P(LN​≤−u)≤2; the event is empty, so nothing is lost.


Retired 2026-10-27: a Lean-vs-paper fidelity audit found this statement weaker than, or narrower than, the cited result. A corrected statement will be resubmitted.

Preamble
import Mathlib
import Definitions.Def_TalagrandConc_ConvexHull_Basic
import Definitions.Def_TalagrandConc_Subsequences_Basic

open MeasureTheory
open scoped ENNReal
Formal statement
namespace TalagrandConc.Subsequences

/-- Talagrand (1995), Theorem 7.1.2, p. 153. Let `X_1, …, X_N` be i.i.d. on `[0,1]` with a
non-atomic law `μ` (e.g. uniform), realised as the coordinates of `[0,1]^N` under
`P = μ^{⊗N}`, and let `M` be a median of `L_N`. For all `u > 0`,
(7.1.3) `P(L_N ≥ M + u) ≤ 2 exp(−u² / (4(M + u)))` and
(7.1.4) `P(L_N ≤ M − u) ≤ 2 exp(−u² / (4M))`. -/
theorem lis_concentration {N : ℕ} (μ : Measure unitInterval) [IsProbabilityMeasure μ]
    (hμ : ∀ t : unitInterval, μ {t} = 0) (M : ℝ)
    (hM : TalagrandConc.BinPacking.IsMedian (Measure.pi fun _ : Fin N => μ) (fun x => (lis x : ℝ)) M)
    (u : ℝ) (hu : 0 < u) :
    (Measure.pi fun _ : Fin N => μ) {x | M + u ≤ (lis x : ℝ)} ≤
        ENNReal.ofReal (2 * Real.exp (-(u ^ 2 / (4 * (M + u))))) ∧
      (Measure.pi fun _ : Fin N => μ) {x | (lis x : ℝ) ≤ M - u} ≤
        ENNReal.ofReal (2 * Real.exp (-(u ^ 2 / (4 * M)))) := by sorry

end TalagrandConc.Subsequences
Source
Talagrand, Concentration of measure and isoperimetric inequalities in product spaces, Publ. Math. IHÉS 81 (1995), p. 153, Theorem 7.1.2, Eqs. (7.1.3)–(7.1.4)
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