Convexity is equivalent to convexity of the restricted epigraph
ProvedVectorSpaceOpt.convexOn_iff_convex_restrictedEpigraphLet be a real vector space, let , and let . Then is convex on if and only if its restricted epigraph is convex:
Here ConvexOn includes convexity of , and the pair order is rather than the scalar-first order printed in the book. This theorem makes the geometric object used by separation arguments explicit and reusable. It is the exact epigraph criterion that connects the elementary convexity section to conjugate and biconjugate constructions. Closedness can later be imposed on this same geometric set.
import Mathlib
namespace VectorSpaceOpt
/-- Luenberger, Chapter 7, §7.8, Proposition 2. -/
theorem convexOn_iff_convex_restrictedEpigraph
{X : Type*} [AddCommGroup X] [Module ℝ X]
(C : Set X) (f : X → ℝ) :
ConvexOn ℝ C f ↔
Convex ℝ {p : X × ℝ | p.1 ∈ C ∧ f p.1 ≤ p.2} := by sorry
end VectorSpaceOptRead-back
What the Lean code literally says, in plain math · gpt-5
For every additive commutative group equipped with a real vector-space structure, every set , and every function , is convex on if and only if the subset is convex over . There is no norm, topology, nonemptiness, or separate closedness hypothesis; may be empty, and values of outside do not enter the displayed set.
Confirmed by the mission captain (proposal self-audit).