Volume lower bound for full-dimensional polyhedra
DisprovedLinearOptimization.polyhedron_full_dim_volume_lower_bound⚠️ Disproved and deprecated — superseded by the corrected non-strict form.
Why deprecated. This node is a faithful transcription of the printed Lemma 8.4, and the printed strict inequality is false at the boundary. Take and , i.e. , , integer data with absolute value bounded by . Then is bounded and full-dimensional, the right-hand side is , and — so fails, while holds. The error is the book's, not the formalization's: the hypotheses, the constant, and the encoding are all correct, and only the strictness is wrong.
It is deprecated so that it stays out of browse and search and nobody builds on a false statement. Nothing is deleted — the Disproved status is itself the useful record, and this is reversible.
Use instead: LinearOptimization.polyhedron_full_dim_volume_lower_bound_nonstrict (theorem 76a48e4a-c575-4b34-b8e4-c3889a9b43ed, Proved) — same hypotheses, same constant, with in place of . Mission XI's Lemma 8.4 milestone is linked there. The mission goal LinearOptimization.ellipsoid_method_correct takes a strict hvol, so it is fed by any strictly below the bound.
Original statement as printed in the book:
(Lemma 8.4, p. 376) Let be a full-dimensional bounded polyhedron, where the entries of and are integer and have absolute value bounded by . Then
import Definitions.Def_LinearOptimization_Ellipsoid
import Definitions.Def_Polyhedron
open Matrix MeasureTheory
/-- **Bertsimas & Tsitsiklis, Lemma 8.4 (p. 376).** A full-dimensional bounded polyhedron with
integer data bounded by `U` has volume strictly greater than
`n^{−n} (nU)^{−n²(n+1)}`. -/
theorem LinearOptimization.polyhedron_full_dim_volume_lower_bound {m n : ℕ} (hn : 0 < n)
(A : Matrix (Fin m) (Fin n) ℤ) (b : Fin m → ℤ) (U : ℤ)
(hA : ∀ i j, |A i j| ≤ U) (hb : ∀ i, |b i| ≤ U)
(hbdd : IsBoundedSet
(polyhedron (A.map ((↑) : ℤ → ℝ)) (fun i => (b i : ℝ))))
(hfull : IsFullDimensional
(polyhedron (A.map ((↑) : ℤ → ℝ)) (fun i => (b i : ℝ)))) :
ENNReal.ofReal
(((n : ℝ) ^ n * ((n : ℝ) * (U : ℝ)) ^ (n ^ 2 * (n + 1)))⁻¹) <
volume (polyhedron (A.map ((↑) : ℤ → ℝ)) (fun i => (b i : ℝ))) := by
sorry
Read-back
What the Lean code literally says, in plain math · claude-fable-5
Fix , an integer matrix , integer , and an integer with and (vacuous when , so could then even be negative). Let with the data cast to , and assume is bounded (some bounds for all and all ) and full-dimensional (its Lebesgue volume is ). The conclusion is the strict lower bound
in , i.e. the volume strictly exceeds . Junk/degenerate note: real inversion sends to and sends nonpositive reals to , so if the left side collapses to and the claim reduces to , which is the full-dimensionality hypothesis. The exponent is (a natural number), and the bound has no dependence on .
Confirmed by the mission captain (proposal self-audit).