Proposition 3.a — the proximal objective ½‖u − z‖² + f(u) has a strict minimum
ProvedMoreauProx.Decomposition.prox_strict_minimumLet be a real Hilbert space, and . Then the function
with values in , has a strict minimum: there is a point such that for every .
The minimizer is therefore unique; Moreau denotes it , the proximal point of relative to (3.b). For the indicator function of a nonempty closed convex set , it is the nearest-point projection onto .
Formalization Note "Strict minimum" is stated literally, as strict inequality at every other point, which gives both existence and uniqueness of the minimizer. The sum is computed in EReal.
import Mathlib import Definitions.Def_MoreauProx_Decomposition_ConvexDuality
namespace MoreauProx.Decomposition
open scoped InnerProductSpace
/-- Moreau 1965, Proposition 3.a, p. 278: for `f ∈ Γ₀(H)` and every `z ∈ H`, the function
`Φ(u) = ½‖u − z‖² + f(u)` has a strict minimum: a point `x` with `Φ(x) < Φ(u)` for all `u ≠ x`. -/
theorem prox_strict_minimum {H : Type*} [NormedAddCommGroup H] [InnerProductSpace ℝ H]
[CompleteSpace H] (f : H → EReal) (hf : GammaZero f) (z : H) :
∃ x : H, ∀ u : H, u ≠ x → proxObjective f z x < proxObjective f z u := by sorry
end MoreauProx.Decomposition
Read-back
What the Lean code literally says, in plain math · claude-opus-5-5
Let be a real Hilbert space. Let belong to , meaning:
- is never ;
- is finite somewhere;
- the epigraph is convex;
- is lower semicontinuous.
Let be arbitrary. Define, in the extended reals,
The theorem asserts that there exists such that
So is a strict global minimiser of . Uniqueness of such an follows from the strict inequality, but it is not stated separately.
Degenerate cases. If , there is no , so the statement holds trivially with . If has a nonzero element, the strict inequality against some forces , and therefore is finite.
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.