Eq. (4) — prox-inequality characterising
ProvedGoldenRatioVI.Explicit.prox_inequalityconvex-analysisp2o-batch-p200ap2o-gran-per-chapterp2o-plan-paperp2o-v1proximal-operator
Let be a finite-dimensional real inner product space and let be proper, convex and lower semicontinuous. For ,
This characteristic property of the proximal operator is used in every convergence proof of the paper; applied to it yields the inequalities (19)–(20) behind the energy estimate (27).
Formalization Note The right-hand side is written without extended-real subtraction as ; it forces (take ), as the left-hand side does.
Preamble
import Mathlib import Definitions.Def_GoldenRatioVI_Explicit_viProblem import Definitions.Def_GoldenRatioVI_Shared_IsProxPoint
Formal statement
namespace GoldenRatioVI.Explicit
/-- Eq. (4) of Malitsky (p. 3), the prox-inequality: for a proper convex lsc
`g : E → (−∞, +∞]`, `xbar = prox_g(w)` iff `⟨xbar − w, x − xbar⟩ ≥ g(xbar) − g(x)` for all `x`,
written without `EReal` subtraction as `g(xbar) ≤ ⟨xbar − w, x − xbar⟩ + g(x)`. -/
theorem prox_inequality {E : Type*} [NormedAddCommGroup E] [InnerProductSpace ℝ E]
[FiniteDimensional ℝ E] (g : E → EReal) (hg : IsProperConvexLSC g) (w xbar : E) :
GoldenRatioVI.Shared.IsProxPoint g w xbar ↔
∀ x : E, g xbar ≤ ((inner ℝ (xbar - w) (x - xbar) : ℝ) : EReal) + g x := by sorry
end GoldenRatioVI.Explicit
Source
Malitsky, Golden Ratio Algorithms for Variational Inequalities, preprint (Optimization Online 6598, 2018), p. 3, Eq. (4)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.