Eq. (2.5) — descent estimate along a run
ProvedCalamaiMore.Convergence.run_descentdescentgradient-projectionp2o-batch-p200ap2o-gran-per-chapterp2o-plan-paperp2o-v1
Let be a nonempty closed convex subset of a finite-dimensional real inner product space , continuously differentiable on , and let be a run of the gradient projection method (2.1)–(2.3) with constants and . Then for every ,
Combined with the sufficient decrease condition (2.1), this bounds the decrease of from below by the squared step length divided by the step.
Preamble
import Mathlib import Definitions.Def_CalamaiMore_Convergence_IsGradientProjectionRun
Formal statement
namespace CalamaiMore.Convergence
/-- Calamai–Moré, Eq. (2.5) (p. 98): along a run of the gradient projection method,
`⟨∇f(x_k), x_k - x_{k+1}⟩ ≥ ‖x_{k+1} - x_k‖² / α_k` for every `k`. -/
theorem run_descent {E : Type*} [NormedAddCommGroup E] [InnerProductSpace ℝ E] [FiniteDimensional ℝ E]
(Ω : Set E) (hΩne : Ω.Nonempty) (hΩc : IsClosed Ω) (hΩcv : Convex ℝ Ω)
(f : E → ℝ) (hfd : ∀ x ∈ Ω, DifferentiableAt ℝ f x) (hfc : ContinuousOn (gradient f) Ω)
(γ₁ γ₂ μ₁ μ₂ : ℝ) (hγ₁ : 0 < γ₁) (hγ₂ : 0 < γ₂)
(hμ₁ : μ₁ ∈ Set.Ioo (0 : ℝ) 1) (hμ₂ : μ₂ ∈ Set.Ioo (0 : ℝ) 1)
(x : ℕ → E) (α : ℕ → ℝ) (hrun : IsGradientProjectionRun f Ω γ₁ γ₂ μ₁ μ₂ x α)
(k : ℕ) :
‖x (k + 1) - x k‖ ^ 2 / α k ≤ inner ℝ (gradient f (x k)) (x k - x (k + 1)) := by sorry
end CalamaiMore.Convergence
Source
Calamai & Moré, Projected gradient methods for linearly constrained problems, Math. Programming 39 (1987), p. 98, Eq. (2.5)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.