Theorem 5.1, p. 135 — n = 2, ρ = 1, γ = ½: the three limiting vertex values coincide, f*₁ = f*₂ = f*₃
ProvedNelderMeadLD.Dim2.theorem_5_1convergencenelder-meadp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1strict-convexity
Let be strictly convex with bounded level sets. Run Algorithm NM on with reflection coefficient , contraction coefficient , and any satisfying (2.1), from a nondegenerate initial triangle . Then the three vertex values converge to one common limit:
i.e. .
This is the first convergence result for the standard two-dimensional Nelder–Mead method; it does not say that is the minimum value of .
Formalization Note The theorem names only and ; and are kept general under (2.1) (§5's standing values are , ), which is at least as strong as the page's statement. Existence of the limits is part of the conclusion.
Preamble
import Mathlib import Definitions.Def_NelderMeadLD_Dim2_Algorithm open Filter Topology
Formal statement
namespace NelderMeadLD.Dim2
theorem theorem_5_1 (f : E 2 → ℝ) (hf : StrictConvexOn ℝ Set.univ f)
(hlev : ∀ μ : ℝ, Bornology.IsBounded {x | f x ≤ μ}) (χ σ : ℝ) (hpar : NelderMeadLD.Conv1D.ParamsOK 1 χ (1 / 2) σ)
(Δ : ℕ → Fin 3 → E 2) (hrun : IsNMRun f 1 χ (1 / 2) σ Δ) (hnd : Nondegenerate (Δ 0)) :
∃ L : ℝ, ∀ i : Fin 3, Tendsto (fun k => f (Δ k i)) atTop (𝓝 L) := by sorry
end NelderMeadLD.Dim2
Source
Lagarias, Reeds, Wright & Wright, Convergence properties of the Nelder–Mead simplex method in low dimensions, SIAM J. Optim. 9 (1998), p. 135, Theorem 5.1
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.