Lemma 4.1, p. 9 — step (4.i) of the adaptive projected reflected gradient method is well-defined
ProvedProjReflGrad.Adaptive.lemma_4_1adaptive-step-sizep2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1variational-inequality
Let , , , and let the previous data of Algorithm 4.2 be a reflection , a weight and a step . Let be the trial reflection and the trial step of (4.1). If , then there is with
This is Lemma 4.1: the choice required in step (4.i) of Algorithm 4.2 is always possible.
Formalization Note The statement is phrased on the data of one step; is the step rule of the setting with the convention . The hypothesis is part of the standing range .
Preamble
import Mathlib import Definitions.Def_ProjReflGrad_Adaptive_Setting
Formal statement
namespace ProjReflGrad.Adaptive
/-- Lemma 4.1 (Malitsky 2015, p. 9): step (4.i) of Algorithm 4.2 is well-defined. If the trial
step `λ(ŷ, 1)` (computed from the previous data `yp, taup, lamp`) is at least `λ_{n-1} = lamp > 0`,
then some `λ'_n ∈ [λ_{n-1}, λ(ŷ, 1)]` satisfies (4.2). -/
theorem lemma_4_1 {H : Type*} [NormedAddCommGroup H] [InnerProductSpace ℝ H] (F : H → H) (α lamBar : ℝ) (hα : 0 ≤ α)
(yp yh : H) (taup lamp : ℝ) (hlamp : 0 < lamp)
(hge : lamp ≤ stepCap F α lamBar yp taup lamp yh 1) :
∃ l ∈ Set.Icc lamp (stepCap F α lamBar yp taup lamp yh 1),
‖l • F yh - lamp • F yp‖ ≤ α * ‖yh - yp‖ := by sorry
end ProjReflGrad.Adaptive
Source
Malitsky, Projected Reflected Gradient Methods for Monotone Variational Inequalities, arXiv:1502.04968v1, p. 9, Lemma 4.1
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.