A box minimum has a nonzero principal-minor denominator certificate
ProvedMurtyKabadi.Reduction.box_minimum_integer_minor_certificateencoding-sizelinear-algebraquadratic-programming
Let be a symmetric integer matrix and let . Suppose is a global minimizer of on this box. There are an index set and an integer such that, with ,
This certificate identifies an integer denominator for the optimal value. Together with an encoding-size bound on principal minors, it gives the quantitative gap in Murty and Kabadi's Lemma 2. The empty principal minor has determinant , so vertex minima and are included.
Formalization Note This is a principal-minor formulation of the denominator argument in the proof of Lemma 2, rather than a separately numbered statement in the paper. The chosen minimizer itself may have irrational coordinates; the certificate concerns its optimal value.
Preamble
import Definitions.Def_MurtyKabadi_Reduction_QuadraticProblems open MurtyKabadi.Reduction
Formal statement
theorem MurtyKabadi.Reduction.box_minimum_integer_minor_certificate
{m : ℕ} (D : Matrix (Fin m) (Fin m) ℤ) (hD : D.IsSymm)
(x : Fin m → ℝ) (hx0 : 0 ≤ x) (hx1 : x ≤ 1)
(hmin : ∀ z : Fin m → ℝ, 0 ≤ z → z ≤ 1 →
Q (D.map (Int.cast : ℤ → ℝ)) x ≤ Q (D.map (Int.cast : ℤ → ℝ)) z) :
∃ S : Finset (Fin m),
(D.submatrix (fun i : S => i.1) (fun i : S => i.1)).det ≠ 0 ∧
∃ a : ℤ, Q (D.map (Int.cast : ℤ → ℝ)) x *
((D.submatrix (fun i : S => i.1) (fun i : S => i.1)).det : ℝ) = (a : ℝ) := by sorrySource
K. G. Murty and S. N. Kabadi, Some NP-complete problems in quadratic and nonlinear programming, Mathematical Programming 39 (1987), pp. 122-123, Lemma 2, program (8) and the denominator argument following (9)-(11). https://public.websites.umich.edu/~murty/np.pdf. Derived auxiliary formulation for this formalization, with the mission's existing encSize convention.