(113), p. 29 — G_f(x, η) − ∇h(x) ∈ ∂g(S_f(x, η))
ProvedNAGFlow.APGM.eq_113gradient-mappingp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1proximal-gradientsubdifferential
Let be a real Hilbert space, with , proper, closed and convex, and . For let and be the composite gradient mapping (111). Then
The inclusion says that is a subgradient-type direction of evaluated in a split way, which is why plays the role of in the nonsmooth scheme.
Formalization Note. is a named point satisfying the prox predicate. is encoded by as in the definitions file. The class of is the standing assumption of (110); the inclusion itself only uses the value .
Preamble
import Mathlib import Definitions.Def_NAGFlow_APGM_Setting
Formal statement
namespace NAGFlow.APGM
/-- Inclusion (113) (Luo & Chen, arXiv:1909.03145v4, §7.2.1, p. 29). Let `h ∈ S^{1,1}_{μ,L}` with
`0 ≤ μ ≤ L < ∞`, `g : V → ℝ ∪ {+∞}` proper, closed and convex (encoded by `(g, D)`), `η > 0` and
`x ∈ V`. If `s = S_f(x, η) = prox_{ηg}(x − η∇h(x))`, then
`G_f(x, η) − ∇h(x) ∈ ∂g(S_f(x, η))`, where `G_f(x, η) = (x − S_f(x, η))/η`. -/
theorem eq_113 {V : Type*} [NormedAddCommGroup V] [InnerProductSpace ℝ V] [CompleteSpace V]
(h : V → ℝ) (gradh : V → V) (μ L : ℝ) (hh : NAGFlow.PredCorr.IsS11 h gradh μ L)
(g : V → ℝ) (D : Set V) (hg : IsProperClosedConvex g D) (η : ℝ) (hη : 0 < η)
(x s : V) (hs : IsSf gradh g D η x s) :
gradMap η x s - gradh x ∈ subdiff g D s := by sorry
end NAGFlow.APGM
Source
Luo & Chen, arXiv:1909.03145v4, §7.2.1, (111) and (113), p. 29
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.