Theorem 1, Step 3 — convergence to the zero limit point
ProvedBregmanPPA.Convergence.step3bregman-functionmonotone-operatorp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1proximal-point
Under the standing hypotheses and either (C1) or (C2) of Theorem 1, suppose a subsequence of the Bregman proximal-point run converges to a zero of . Then the entire run converges to that point:
This rules out a second limit point and completes the convergence half of Theorem 1 once Step 2 has identified a zero limit point.
Formalization Note The subsequence indices are strictly increasing. The conclusion is convergence of the whole sequence.
Preamble
import Mathlib import Definitions.Def_BregmanPPA_Convergence_Model open ThreeOpSplitting.Convergence InertialFB.IFB Filter Topology
Formal statement
namespace BregmanPPA.Convergence
variable {H : Type*} [NormedAddCommGroup H] [InnerProductSpace ℝ H] [FiniteDimensional ℝ H]
/-- Theorem 1, proof, Step 3: a zero attained as a limit point attracts the whole run. -/
theorem step3 (T : H → Set H) (S : Set H) (h : H → ℝ) (c : ℕ → ℝ) (x : ℕ → H)
(hT : IsMaximalMonotone T) (hh : IsBregmanFunction S h)
(hdom : dom T ⊆ closure S) (hc : ∀ k, 0 < c k)
(hcinf : ∃ ε : ℝ, 0 < ε ∧ ∀ k, ε ≤ c k)
(hx : IsBregmanPPARun S h T c x)
(hC : closure (dom T) ⊆ S ∨ ∃ f : H → EReal,
IsProperFn f ∧ IsConvexFn f ∧ LowerSemicontinuous f ∧ T = subdiffOp f)
(x' : H) (φ : ℕ → ℕ) (hφ : StrictMono φ)
(hlim : Tendsto (x ∘ φ) atTop (𝓝 x')) (hzero : x' ∈ zer T) :
Tendsto x atTop (𝓝 x') := by sorry
end BregmanPPA.Convergence
Source
Eckstein, Nonlinear proximal point algorithms using Bregman functions, with applications to convex programming, Math. Oper. Res. 18(1) (1993), p. 208, Theorem 1 proof, Step 3, https://doi.org/10.1287/moor.18.1.202
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.