Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Theorem 1.2 — Euclidean coreset lower bound at every error scale

Proved
TightCoreset2027.Euclidean.euclidean_coreset_lower_bound

by mikedeng1 · Nov 9, 2026 · Mathlib 0df444a (Lean v4.33.1)

clusteringcomputational-geometrycoresetsp2o-batch-stoc27ap2o-gran-per-chapterp2o-plan-paperp2o-v1

Fix any real exponent z≥1z\ge1z≥1. There are a positive constant czc_zcz​ and nonnegative integer degrees az,bza_z,b_zaz​,bz​, depending only on zzz, such that for every integer k≥1k\ge1k≥1 and every 0<ε<10<\varepsilon<10<ε<1, some positive finite dimension ddd and finite set X⊆RdX\subseteq\mathbb R^dX⊆Rd force every strong ε\varepsilonε-coreset (S,w)(S,w)(S,w) for (k,z)(k,z)(k,z)-clustering to satisfy

∣S∣≥czmin⁡{k(2z+2)/(z+2)ε−2, kε−z−2}[log⁡(2+k)]az[log⁡(2+1/ε)]bz.|S|\ge\frac{c_z\min\{k^{(2z+2)/(z+2)}\varepsilon^{-2},\ k\varepsilon^{-z-2}\}}{[\log(2+k)]^{a_z}[\log(2+1/\varepsilon)]^{b_z}}.∣S∣≥[log(2+k)]az​[log(2+1/ε)]bz​cz​min{k(2z+2)/(z+2)ε−2, kε−z−2}​.

The coreset must preserve every set of exactly kkk distinct ambient centers, with absolute relative error, and may use any nonnegative weights on any subset of XXX. No sampling or mass constraint is imposed. The dimension and dataset may depend on z,k,εz,k,\varepsilonz,k,ε but precede the candidate subset and its weights. This is the full lower bound in Theorem 1.2 under footnote 1's asymptotic convention, including all real exponents, both error regimes, k=1k=1k=1, and arbitrarily small positive errors.

Formalization Note The constants and log degrees are quantified before k,εk,\varepsilonk,ε. All logarithm arguments exceed 222 in this domain. The calibrated general-power construction and boundary arguments of §2.5 remain proof obligations; they are not hypotheses of this theorem.

Differences from the source:

  • encoding: tilde-Omega with z-only constants and finite polylogarithmic factors by footnote 1 → positive c_z and natural degrees a_z,b_z chosen before k and epsilon, with the displayed denominator; reason: this is precisely the source asymptotic convention, not instance-dependent constants
  • encoding: a finite subset of Euclidean space and every exact-k ambient center set → Finset (EuclideanSpace R (Fin d)) for data and all ambient queries; reason: finite sets preserve distinct support and exact cardinality; Euclidean coordinates use an isometric representation
  • encoding: arbitrary nonnegative weights defined on S → an arbitrary ambient NNReal function whose values outside S are ignored; reason: every source weight function extends by zero, including zero weights and the empty candidate
Preamble
import Definitions.Def_TightCoreset2027_Euclidean_coresetModel
set_option autoImplicit false
Formal statement
namespace TightCoreset2027.Euclidean

theorem euclidean_coreset_lower_bound (z : ℝ) (hz : 1 ≤ z) :
    ∃ c : ℝ, 0 < c ∧ ∃ a b : ℕ,
      ∀ k : ℕ, 1 ≤ k → ∀ ε : ℝ, 0 < ε → ε < 1 →
        ∃ d : ℕ, 1 ≤ d ∧ ∃ X : Finset (Point d),
          ∀ (S : Finset (Point d)) (w : Point d → NNReal),
            IsStrongCoreset z ε k X S w →
              c * lowerBoundScale z k ε /
                ((Real.log (2 + (k : ℝ))) ^ a * (Real.log (2 + 1 / ε)) ^ b)
                ≤ (S.card : ℝ) := by sorry

end TightCoreset2027.Euclidean
Source
Junqiu Hu, Lingxiao Huang, Jian Li, Baoduo Xu, STOC 2027 submission #879 (October 2026), Theorem 1.2, printed p. 3 / PDF p. 5; footnote 1 and Definition 1.1, printed p. 1 / PDF p. 3; §2.5, printed pp. 19–21 / PDF pp. 21–23; https://github.com/sigact-stoc/stoc2027/releases/download/papers-1/879.pdf
Human review
  • Endorsed by Shuze Chen · Nov 9, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Nov 9, 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