Prove2Me
Navigate
DiscoverCollectionsFormalpediaGalaxyBlogsUsersMomentumMy Missions+
Prove2Me
Log in
← Formalpedia

Lemma 27, p. 23 — moving a point towards the optimum without changing CMP's output strictly worsens a ratio above 1

Proved
LearnAugFL.CMP.lemma_27

by mikedeng1 · Oct 22, 2026 · Mathlib 0df444a (Lean v4.33.1)

approximation-ratiocoordinatewise-medianfacility-locationlearning-augmentedp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1

Throughout, R2\mathbb R^2R2 carries the Euclidean distance ddd; P=(p1,…,pn)P=(p_1,\dots,p_n)P=(p1​,…,pn​) is a profile of n≥1n\ge 1n≥1 preferred locations; c∈[0,1)c\in[0,1)c∈[0,1) is a confidence parameter with cncncn an integer; Cu(f,P)=1n∑i=1nd(f,pi)C^u(f,P)=\frac1n\sum_{i=1}^n d(f,p_i)Cu(f,P)=n1​∑i=1n​d(f,pi​) is the utilitarian social cost; o(P)o(P)o(P) is an optimal (cost-minimizing) location; and f(P,o^,c)f(P,\hat o,c)f(P,o^,c) is the location chosen by the Coordinatewise Median with Predictions (CMP): each coordinate is the lower median of the n+cnn+cnn+cn values formed by the agents' coordinates and cncncn copies of the prediction's coordinate. The ratio is r(P,o^,c)=Cu(f(P,o^,c),P)/Cu(o(P),P)r(P,\hat o,c)=C^u(f(P,\hat o,c),P)/C^u(o(P),P)r(P,o^,c)=Cu(f(P,o^,c),P)/Cu(o(P),P).

Fix c∈[0,1)c\in[0,1)c∈[0,1), a prediction o^\hat oo^, a profile PPP with optimal location ooo, an agent iii with pi≠op_i\neq opi​=o, and pi′=pi+s(o−pi)p'_i=p_i+s(o-p_i)pi′​=pi​+s(o−pi​) with s∈(0,1]s\in(0,1]s∈(0,1] (a point of the segment (pi,o](p_i,o](pi​,o]). If CMP's output does not change, f(P,o^,c)=f((P−i,pi′),o^,c)f(P,\hat o,c)=f((P_{-i},p'_i),\hat o,c)f(P,o^,c)=f((P−i​,pi′​),o^,c), and r(P,o^,c)>1r(P,\hat o,c)>1r(P,o^,c)>1, then

r((P−i,pi′),o^,c)>r(P,o^,c),r\big((P_{-i},p'_i),\hat o,c\big)>r(P,\hat o,c),r((P−i​,pi′​),o^,c)>r(P,o^,c),

where both ratios are measured against ooo (which remains optimal for the new profile by Lemma 26).

The lemma drives every "move a point towards the optimum" step of the worst-case reduction.

Formalization Note The optimal location o(P)o(P)o(P) is taken as any ooo with IsOpt P o (a minimizer of Cu(⋅,P)C^u(\cdot,P)Cu(⋅,P); it is unique unless the points are collinear). The confidence enters through Admissible c n k: n≥1n\ge1n≥1 and k=cnk=cnk=cn copies of the prediction. Ratios are real quotients ratio P ô o k; on every family used here the optimal cost is positive (if all points sat at o(P)o(P)o(P), CMP would return o(P)o(P)o(P) because cn<ncn<ncn<n), so no ratio is a junk 0/00/00/0. The hypothesis pi≠op_i\ne opi​=o makes the segment (pi,o](p_i,o](pi​,o] nonempty, as the page presupposes.

Preamble
import Mathlib
import Definitions.Def_LearnAugFL_CMP_Setting
Formal statement
namespace LearnAugFL.CMP

theorem lemma_27 (c : ℝ) (hc0 : 0 ≤ c) (hc1 : c < 1) {n k : ℕ} (hnk : Admissible c n k)
    (P : Fin n → Pt) (ô o : Pt) (hP : IsOpt P o) (i : Fin n) (hi : P i ≠ o)
    (s : ℝ) (hs0 : 0 < s) (hs1 : s ≤ 1)
    (hf : cmp P ô k = cmp (Function.update P i (P i + s • (o - P i))) ô k)
    (hr : 1 < ratio P ô o k) :
    ratio P ô o k < ratio (Function.update P i (P i + s • (o - P i))) ô o k := by sorry
end LearnAugFL.CMP
Source
Agrawal, Balkanski, Gkatzelis, Ou & Tan, Learning-Augmented Mechanism Design: Leveraging Predictions for Facility Location, arXiv:2204.01120v1, Lemma 27, p. 23
Human review
  • Endorsed by Shuze Chen · Oct 23, 2026

    Confirmed by the moderator at approval.

  • Endorsed by mikedeng1 · Oct 23, 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