The natural power-cost relaxation
DefinitionKMeansTaper_Supplement_lpPowerapproximation-algorithmsfacility-locationk-meanslinear-programmingp2o-gran-per-chapterp2o-plan-paperp2o-v1
For a metric -means instance and a real exponent , is the optimum of the standard facility-location relaxation
subject to , for every client , and . It is the LP benchmark in Theorem 3.6, and at it is exactly the published KMeansTaper.lp2.
Formalization Note The value is represented as the real infimum of the nonempty, nonnegative feasible objective set. Real powers use Real.rpow; the constraints are exactly those pinned by the base mission's squared LP.
Definition code
import Mathlib
import Definitions.Def_MetricKMedian
import Definitions.Def_KMeansTaper_lp2
namespace KMeansTaper
namespace Supplement
open OAI.MetricKMedian (Instance)
/-- **The power-cost LP value `LP_p`**, introduced immediately before Theorem 3.6, p.5
(PDF p.8): the optimum of the Section 2 relaxation with objective
`∑_{j ∈ J} ∑_{i ∈ F} d(j,i)^p x_{ji}` and facility budget `k`.
Zhu, *A tight approximation for metric k-means via squared tapered rounding*, STOC 2027
submission (manuscript sha256 8b6610229e43dc2441325b120437136afe762dbd6e873f73919fcb6b043fc5a3).
**Formalization Note.** This uses the same standard relaxation pinned by the published `lp2`:
`0 ≤ x_{ji} ≤ y_i ≤ 1`, every client is assigned one unit, and `∑_i y_i ≤ k`.
The real infimum is over a nonempty set bounded below by zero. Real powers use `Real.rpow`. -/
noncomputable def lpPower (I : Instance) (p : ℝ) : ℝ :=
sInf {v : ℝ | ∃ (x : Fin I.n → Fin I.n → ℝ) (y : Fin I.n → ℝ),
(∀ j i, 0 ≤ x j i ∧ x j i ≤ y i) ∧ (∀ i, y i ≤ 1) ∧
(∀ j ∈ I.clients, ∑ i ∈ I.facilities, x j i = 1) ∧ ∑ i ∈ I.facilities, y i ≤ I.k ∧
v = ∑ j ∈ I.clients, ∑ i ∈ I.facilities, (I.distance j i : ℝ) ^ p * x j i}
/-- At exponent two, `lpPower` is exactly the published squared relaxation `lp2`. -/
@[simp] theorem lpPower_two (I : Instance) : lpPower I 2 = lp2 I := by
simp [lpPower, lp2]
end Supplement
end KMeansTaper
Source
Zhu, A tight approximation for metric k-means via squared tapered rounding, STOC 2027 submission (manuscript sha256 8b6610229e43dc2441325b120437136afe762dbd6e873f73919fcb6b043fc5a3), Section 2 ("Problem"), p.3 (PDF p.6), and paragraph before Theorem 3.6, p.5 (PDF p.8)