Proposition 13.22 — as a projection, for polymatroid rank functions
OpenDisjunctive.Polymatroids.pi_projection_characterization_v2blockerdisjunctive-programmingpolymatroidsprojectionunion-of-polytopes
This is Proposition 13.22 of Balas's Disjunctive Programming.
Let be polymatroid rank functions (, nondecreasing, submodular), , and . Then
Formalization Note. The retired version had no hypotheses on . For arbitrary set functions there need not be a single multiplier vector that works for both functions, and the equality fails: with , , , , . The new statement adds the standing assumption of §13.8 that both are polymatroid rank functions (IsPolymatroidRankFunction, already in the definition module). Under it, the greedy dual solution depends only on the order of and therefore serves both functions at once.
Preamble
import Mathlib import Definitions.Def_Disjunctive_Polymatroids_Basic
Formal statement
namespace Disjunctive.Polymatroids
/-- Proposition 13.22 (Balas, *Disjunctive Programming*, Springer 2018, §13.8, p. 231): for
polymatroid rank functions `r₁, r₂` on `N` (the standing assumption of §13.8: `r(∅) = 0`,
nondecreasing, submodular), `Π := {π ≥ 0 : πx ≤ 1 for x ∈ P(r₁) ∪ P(r₂)}` equals the projection
onto `π` of `{(π,u) : π_j ≤ Σ_{A∋j} u_A (j ∈ N), Σ_A u_A r_i(A) ≤ 1 (i = 1,2), π, u ≥ 0}`.
Correction w.r.t. the retired version: `r₁, r₂` were arbitrary set functions; the hypotheses
`hr1`, `hr2` now require them to be polymatroid rank functions, as the section assumes. -/
theorem pi_projection_characterization_v2 {n : ℕ} (r1 r2 : Finset (Fin n) → ℝ)
(hr1 : IsPolymatroidRankFunction r1) (hr2 : IsPolymatroidRankFunction r2) :
PiSet r1 r2 = PiProjSystem r1 r2 := by sorry
end Disjunctive.Polymatroids
Source
E. Balas, Disjunctive Programming, Springer 2018, DOI 10.1007/978-3-030-00148-3, §13.8, p. 231, Proposition 13.22