Proposition 2.9 -- conjugate quadratic forms correspond to inverse matrices
ProvedDiscreteConvex.CombinatorialB.prop_2_9_quadratic_conjugate_iff_inverseconvex-optimizationdiscrete-convex-analysis
Let be positive-definite symmetric matrices. and are Legendre-Fenchel conjugate to each other if and only if and are inverse to each other.
(Murota, Discrete Convex Analysis, SIAM 2003, DOI 10.1137/1.9780898718508, p.68, Proposition 2.9.)
Preamble
import Mathlib import Definitions.Def_DiscreteConvex_CombinatorialB_QF import Definitions.Def_DiscreteConvex_CombinatorialB_Conjugate
Formal statement
/-!
Murota, *Discrete Convex Analysis*, SIAM 2003, p.68, Proposition 2.9, in
`DiscreteConvex.CombinatorialB`.
-/
namespace DiscreteConvex.CombinatorialB
/-- **Proposition 2.9.** Let `M, L` be positive-definite symmetric matrices. The quadratic
forms `f(x) = (1/2)x⊤Mx` and `g(p) = (1/2)p⊤Lp` are conjugate to each other with respect to
the Legendre-Fenchel transformation (1.6) if and only if `M` and `L` are inverse to each other. -/
theorem prop_2_9_quadratic_conjugate_iff_inverse {V : Type*} [Fintype V] [DecidableEq V]
(M L : Matrix V V ℝ) (hMsymm : M.IsSymm) (hLsymm : L.IsSymm) (hMpd : M.PosDef)
(hLpd : L.PosDef) :
((∀ p : V → ℝ, Conjugate (QF M) p = ((QF L p : ℝ) : EReal)) ∧
(∀ x : V → ℝ, Conjugate (QF L) x = ((QF M x : ℝ) : EReal))) ↔ M * L = 1 := by sorry
end DiscreteConvex.CombinatorialB
Source
Murota, Discrete Convex Analysis, SIAM 2003, DOI 10.1137/1.9780898718508, p.68, Proposition 2.9
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.