Proposition 2, pp. 21–22 — the multiplicity bound k ≥ aff dim{b(γ) : γ ∈ ℱ} + 1 of Theorems 1 and 2 cannot be lowered to k ≥ aff dim{b(γ) : γ ∈ ℱ}
ProvedQCQPTightness.Sharp.proposition_2convex-hullcounterexamplep2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1qcqpquadratic-eigenvalue-multiplicitysdp-relaxation
For any positive integers and there exists a QCQP in variables with constraints such that
- Assumptions 1 and 3 are satisfied,
- the quadratic eigenvalue multiplicity of the QCQP is ,
- for every semidefinite face of ,
- but , and hence .
Theorems 1 and 2 of the paper guarantee and under Assumptions 1 and 3 when for every semidefinite face. This proposition shows that the "+1" cannot be removed.
Formalization Note The QCQP is existentially quantified over QCQP (n * k) (k + 1), so and are fixed as on the page. "The multiplicity is " is IsQuadEigMult k ( is the largest admissible integer). Faces are nonempty. Both conclusions of the last bullet are stated.
Preamble
import Mathlib import Definitions.Def_QCQPTightness_Sharp_QCQP import Definitions.Def_QCQPTightness_Sharp_Faces import Definitions.Def_QCQPTightness_Sharp_Mult
Formal statement
namespace QCQPTightness.Sharp
/-- Proposition 2 (arXiv:1911.09195v3, pp. 21–22): for all positive integers `n`, `k` there is a
QCQP in `N = nk` variables with `m = k + 1` constraints satisfying Assumptions 1 and 3, with
quadratic eigenvalue multiplicity `k`, with `k ≥ aff dim{b(γ) : γ ∈ ℱ}` for every semidefinite
face `ℱ` of `Γ`, but with `Opt ≠ Opt_SDP` (and hence `conv(𝒟) ≠ 𝒟_SDP`). -/
theorem proposition_2 : ∀ n k : ℕ, 0 < n → 0 < k →
∃ P : QCQP (n * k) (k + 1),
P.Assumption1 ∧ P.Assumption3 ∧ P.IsQuadEigMult k ∧
(∀ F : Set (Fin (k + 1) → ℝ), P.IsSemidefiniteFace F → affdim (P.bγ '' F) ≤ k) ∧
P.Opt ≠ P.OptSDP ∧ convexHull ℝ P.D ≠ P.DSDP := by sorry
end QCQPTightness.Sharp
Source
arXiv:1911.09195v3, Proposition 2, pp. 21–22
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.