§5, p. 220 — (x*, p*) is a saddle point of l iff (0, 0) ∈ K(x*, p*)
ProvedBregmanPPA.ProxMult.saddle_iff_zerolagrangianmonotone-operatorp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1saddle-point
For a problem of the form (10) with Lagrangian and saddle operator , and any ,
The zeros of are therefore exactly the saddle pairs, which the paper identifies with the optimal solution–Lagrange multiplier pairs of (10). This translates the conclusion of Theorem 1 for into the conclusion of Theorem 8.
Formalization Note The page states two equivalences: " optimal and a Lagrange multiplier" iff the saddle inequality, iff . This item states the second. The first is how the paper names saddle pairs; the mission takes the saddle inequality itself as the meaning of "optimal solution–Lagrange multiplier pair".
Preamble
import Mathlib import Definitions.Def_BregmanPPA_Convergence_Model import Definitions.Def_BregmanPPA_IneqMult_Program import Definitions.Def_BregmanPPA_ProxMult_Saddle open ThreeOpSplitting.Convergence
Formal statement
namespace BregmanPPA.ProxMult
/-- §5, p. 220: `(x*, p*)` satisfies the saddle inequality
`l(x*, p) ≤ l(x*, p*) ≤ l(x, p*)` for all `x, p` if and only if `(0, 0) ∈ K(x*, p*)`. -/
theorem saddle_iff_zero {n m : ℕ} (C : Set (BregmanPPA.IneqMult.E n)) (f : BregmanPPA.IneqMult.E n → EReal) (g : Fin m → BregmanPPA.IneqMult.E n → EReal)
(hP : BregmanPPA.IneqMult.IsConvexProgram C f g) (x' : BregmanPPA.IneqMult.E n) (p' : BregmanPPA.IneqMult.E m) :
IsSaddlePair C f g x' p' ↔ pair x' p' ∈ zer (opK C f g) := by sorry
end BregmanPPA.ProxMult
Source
Eckstein, Nonlinear proximal point algorithms using Bregman functions, with applications to convex programming, Math. Oper. Res. 18(1) (1993), p. 220, §5 (saddle inequality ⇔ (0,0) ∈ K(x*, p*))
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.