§2.3, p. 130 — minimizers of f are exactly fixed points of prox_{λf}
ProvedProxAlg.FixedPoint.fixed_point_iff_minimizerconvex-optimizationfixed-pointp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1proximal-operator
Let be closed, proper and convex, and let . A point minimizes if and only if it is a fixed point of the proximal operator of :
This identifies the optimization problem with a fixed-point problem. The paper prints the case and explicitly observes that positive scaling leaves minimizers unchanged.
Formalization Note The function is the extended-real pointwise product, with ensuring that .
Preamble
import Mathlib import Definitions.Def_MoreauProx_Decomposition_ConvexDuality import Definitions.Def_ProxAlg_FixedPoint_Basic
Formal statement
namespace ProxAlg.FixedPoint
theorem fixed_point_iff_minimizer {n : ℕ} (f : EuclideanSpace ℝ (Fin n) → EReal)
(hf : MoreauProx.Decomposition.GammaZero f) (lam : ℝ) (hlam : 0 < lam)
(xs : EuclideanSpace ℝ (Fin n)) :
(∀ y, f xs ≤ f y) ↔
MoreauProx.Decomposition.IsProx (fun u => ((lam : ℝ) : EReal) * f u) xs xs := by sorry
end ProxAlg.FixedPoint
Source
Parikh & Boyd, Proximal Algorithms, Found. Trends Optim. 1(3) (2014), §2.3, p. 130, opening fixed-point statement and parenthesis
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.