Theorem 23 — stability of γ–tilted secant convexity under limits
OpenTiltedCvx.Supplement.theorem_23convex-analysisgeneralized-convexitynonsmooth-analysisoptimizationp2o-gran-per-chapterp2o-plan-bookp2o-strengthenedp2o-v1
Let be a non-empty convex set and . Let be proper –tilted secant convex functions that converge pointwise to finite values on a dense subset . Then there is a continuous function such that
- uniformly on compact subsets of ;
- is –tilted secant convex.
Moreover, if there is a function with pointwise on , then never takes the value and is –tilted secant convex.
No lower semicontinuity is assumed.
Formalization Note The secant convexity of and of is that of their proper extensions (properExt (ri K) f, and on , off ).
Differences from the source:
- generalization: functions on ℝ^d → functions on a finite-dimensional real inner product space E; reason: every such E is isometric to ℝ^d and ℝ^d = EuclideanSpace ℝ (Fin d) is an instance, so the statement specializes to the paper's
- encoding: ℝ̄ = ℝ ∪ {±∞}, dom(f) = {f finite} → EReal and the published TiltedCvx.effDom; reason: Lean convention; same instances
- encoding: D is a dense subset of K → D ⊆ K ⊆ closure D; reason: the meaning of density in K
- encoding: fₙ → f uniformly on compact subsets of ri(K) → for every compact C ⊆ ri K and ε > 0, eventually f x − ε ≤ fₙ x ≤ f x + ε on C (in EReal); reason: uniform convergence of ℝ̄-valued fₙ to a real f; it forces fₙ to be finite on C eventually, as |fₙ − f| ≤ ε requires
- encoding: a continuous f : ri(K) → ℝ, and g : K → ℝ̄ → f : E → ℝ continuous on ri K, its secant convexity that of properExt (ri K) f; g : E → EReal read on K, extended by ⊤ off K; reason: the paper's convention that class membership refers to the proper extension (§3)
Preamble
import Mathlib import Definitions.Def_TiltedCvx_IsProper import Definitions.Def_TiltedCvx_TiltedSecant import Definitions.Def_TiltedCvx_properExt open Filter Topology Classical
Formal statement
namespace TiltedCvx.Supplement
/-- **Theorem 23 (stability of γ–tilted secant convexity under limits).**
Pedramfar–Khanna–Quinn–Aggarwal, arXiv:2611.03472, Theorem 23, p.28. Let `K ⊆ ℝ^d` be a non-empty
convex set and `γ ∈ (0, 1]`. Let `fₙ : ℝ^d → ℝ̄` be proper γ–tilted secant convex functions that
converge pointwise to finite values on a dense subset `D ⊆ K`. Then there is a continuous function
`f : ri(K) → ℝ` such that (i) `fₙ → f` uniformly on compact subsets of `ri(K)`; (ii) `f` is
γ–tilted secant convex. Moreover, if there is a function `g : K → ℝ̄` such that `fₙ → g` pointwise
on `K`, then `g` never takes the value `−∞` and is γ–tilted secant convex.
**Formalization Note** No lower semicontinuity is assumed. "`D` dense in `K`" is
`D ⊆ K ⊆ closure D`. `f : ri(K) → ℝ` is a real function on `E` read on `ri(K)`, and its secant
convexity is that of its proper extension (`properExt (ri K) f`). Uniform convergence on a compact
`C ⊆ ri(K)`: for every `ε > 0`, eventually `f x − ε ≤ fₙ x ≤ f x + ε` on `C`. `g : K → ℝ̄` is a
function on `E` read on `K`; its secant convexity is that of `g` on `K`, `+∞` off `K`. -/
theorem theorem_23 {E : Type*} [NormedAddCommGroup E] [InnerProductSpace ℝ E]
[FiniteDimensional ℝ E] (K : Set E) (hKne : K.Nonempty) (hKc : Convex ℝ K)
(γ : ℝ) (hγ0 : 0 < γ) (hγ1 : γ ≤ 1) (fn : ℕ → E → EReal)
(hfn : ∀ n, TiltedCvx.IsProper (fn n) ∧ TiltedCvx.TiltedSecant γ (fn n))
(D : Set E) (hDK : D ⊆ K) (hKD : K ⊆ closure D)
(hconv : ∀ x ∈ D, ∃ L : ℝ, Tendsto (fun n => fn n x) atTop (𝓝 (L : EReal))) :
∃ f : E → ℝ, ContinuousOn f (intrinsicInterior ℝ K) ∧
(∀ C ⊆ intrinsicInterior ℝ K, IsCompact C → ∀ ε : ℝ, 0 < ε →
∀ᶠ n in atTop, ∀ x ∈ C,
(((f x - ε : ℝ) : EReal) ≤ fn n x ∧ fn n x ≤ ((f x + ε : ℝ) : EReal))) ∧
TiltedCvx.TiltedSecant γ (TiltedCvx.properExt (intrinsicInterior ℝ K) f) ∧
(∀ g : E → EReal, (∀ x ∈ K, Tendsto (fun n => fn n x) atTop (𝓝 (g x))) →
(∀ x ∈ K, g x ≠ ⊥) ∧
TiltedCvx.TiltedSecant γ (fun x => if x ∈ K then g x else ⊤)) := by sorry
end TiltedCvx.Supplement
Source
Pedramfar–Khanna–Quinn–Aggarwal, Tilted Convexity and an Exact Offline–Online Separation, arXiv:2611.03472, Theorem 23, p.28