Remarque 1 of §IV — for uniformly convex f the proximal iterates converge strongly to the unique solution of (12)
ProvedMartinetReg.ConvexMin.remark1Assume the hypotheses of §4.1 ( a real Hilbert space, nonempty, closed and convex, convex and lower semicontinuous with bounded sublevel sets ), and assume moreover that is uniformly convex: there is with for all such that
for all and with . Let be generated by (13) from some . Then (12) has a unique solution , and
that is, converges strongly to .
The remark upgrades Théorème 3 from weak cluster points to strong convergence of the whole sequence when has a modulus of uniform convexity.
Formalization Note Uniform convexity is Mathlib's UniformConvexOn Set.univ φ f; with only this would be plain convexity, so the hypothesis for is what makes it uniform convexity. No monotonicity of is assumed. All hypotheses of §4.1 are kept, since the remark is made under them. "Converge fortement" is convergence in the norm topology.
import Mathlib import Definitions.Def_MartinetReg_ConvexMin_Setting
namespace MartinetReg.ConvexMin
open Filter Topology
/-- Remarque 1 of §IV, p. 158: if moreover f is uniformly convex, problem (12) has a unique
solution and the sequence {xⁿ} generated by (13) converges strongly (in norm) to it. -/
theorem remark1 {H : Type*} [NormedAddCommGroup H] [InnerProductSpace ℝ H]
[CompleteSpace H] (f : H → ℝ) (C : Set H) (hS : Standing f C)
(φ : ℝ → ℝ) (hφ : ∀ t : ℝ, 0 < t → 0 < φ t) (hunif : UniformConvexOn Set.univ φ f)
(x : ℕ → H) (hx : IsProxSeq f C x) :
∃ xs : H, solSet f C = {xs} ∧ Tendsto x atTop (𝓝 xs) := by sorry
end MartinetReg.ConvexMin
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.