Proof of Theorem 6 — when , and every limit point of solves (7)
ProvedBregmanPPA.EqMult.theorem6_limit_pointsAssume the setting of Theorem 6: problem (7) with closed convex, convex and lower semicontinuous, , the dual functional not identically ; a Bregman function with zone and ; scalars with ; and conforming to recursion (9). Suppose that , where is an optimal multiplier (a maximizer of ), and let be a limit point of , i.e. the limit of some subsequence . Then
- ;
- ;
- , where ;
- solves (7): , , and for every with .
This is the primal half of Theorem 6: once the multipliers converge to a Lagrange multiplier, the primal iterates are asymptotically feasible and all their limit points are optimal, even though itself need not converge.
Formalization Note A limit point is a subsequential limit: a strictly increasing with . is the adjoint of the linear map . The subdifferential is that of the EReal function . The hypotheses are the standing assumptions of Theorem 6 together with the convergent case considered in its proof.
import Mathlib import Definitions.Def_InertialFB_IFB_ConvexAnalysis import Definitions.Def_fenchelConjugate import Definitions.Def_BregmanPPA_Convergence_Model import Definitions.Def_BregmanPPA_EqMult_EqConstrainedProgram open InertialFB.IFB ConvexOptimization Filter Topology
namespace BregmanPPA.EqMult
theorem theorem6_limit_points {n m : ℕ}
(A : EuclideanSpace ℝ (Fin n) →L[ℝ] EuclideanSpace ℝ (Fin m))
(b : EuclideanSpace ℝ (Fin m)) (X : Set (EuclideanSpace ℝ (Fin n)))
(f : EuclideanSpace ℝ (Fin n) → EReal)
(hP : IsEqProgram X f) (hd : ∃ p, dualEq A b X f p ≠ ⊥)
(h : EuclideanSpace ℝ (Fin m) → ℝ) (hh : BregmanPPA.Convergence.IsBregmanFunction Set.univ h)
(him : Function.Surjective (gradient h))
(c : ℕ → ℝ) (hc : ∀ k, 0 < c k) (hcinf : ∃ ε > 0, ∀ k, ε ≤ c k)
(x : ℕ → EuclideanSpace ℝ (Fin n)) (p : ℕ → EuclideanSpace ℝ (Fin m))
(hrun : IsEqMultRun A b X f h c x p)
(pstar : EuclideanSpace ℝ (Fin m)) (hpstar : IsOptimalMultiplier A b X f pstar)
(hp : Tendsto p atTop (𝓝 pstar))
(φ : ℕ → ℕ) (hφ : StrictMono φ) (x' : EuclideanSpace ℝ (Fin n))
(hx' : Tendsto (x ∘ φ) atTop (𝓝 x')) :
Tendsto (fun k => A (x k)) atTop (𝓝 b) ∧ A x' = b ∧
-(ContinuousLinearMap.adjoint A pstar) ∈ BregmanPPA.Convergence.subdiffOp (fX X f) x' ∧
IsSolution7 A b X f x' := by sorry
end BregmanPPA.EqMult
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.