VectorSpaceOpt.ConeConvexOn
DefinitionVectorSpaceOpt_coneConvexOnLet be a subset of a real vector space , let be a convex cone in , and let . The predicate ConeConvexOn P Ω G requires to be convex and, for all and nonnegative with , requires
This is Luenberger's vector-valued convexity with respect to the cone order, not componentwise convexity in a chosen basis. The definition packages exactly the condition needed to make feasible sets under convex combinations and the perturbation value convex, and it remains reusable for conic programs in non-lattice spaces.
import Definitions.Def_VectorSpaceOpt_coneLE
namespace VectorSpaceOpt
/-- Convexity of a vector-valued map with respect to a cone order. -/
def ConeConvexOn
{X Z : Type*}
[AddCommGroup X] [Module ℝ X]
[AddCommGroup Z] [Module ℝ Z]
(P : ConvexCone ℝ Z) (Ω : Set X) (G : X → Z) : Prop :=
Convex ℝ Ω ∧
∀ ⦃x₀⦄, x₀ ∈ Ω → ∀ ⦃x₁⦄, x₁ ∈ Ω →
∀ ⦃a b : ℝ⦄, 0 ≤ a → 0 ≤ b → a + b = 1 →
coneLE P (G (a • x₀ + b • x₁)) (a • G x₀ + b • G x₁)
end VectorSpaceOptRead-back
What the Lean code literally says, in plain math · gpt-5
For additive commutative groups and equipped with real vector-space structures, a real convex cone in , a set , and a map , this defines to be cone-convex on when is convex and, for every and all real with , the vector belongs to . Equivalently, the definition applies the custom cone relation exactly when to compare with . There is no topology or norm requirement; if is empty, its convexity and the quantified comparison are both vacuous.
Confirmed by the mission captain (proposal self-audit).