Sums of finitely many convex sets #
A sum C₁ + ⋯ + Cₘ of subsets of E is the image of the product set ∏ Cᵢ ⊆ ι → E under the
sum map (xᵢ) ↦ ∑ xᵢ, so every question about the closure and the recession cone of a finite sum
becomes a question about a linear image, which the recession calculus of a linear map answers.
Convex.isClosed_sum, Convex.closure_sum_eq and Convex.recessionCone_sum are the m-ary
statements: closure and recession cone both distribute over a finite sum of convex sets.
The cancellation hypothesis is genuinely m-ary: a sum of m directions of recession can vanish
without any two of them cancelling, so it is stated on the family, and the two-set results of
Recession/Closedness.lean are kept separately rather than derived from these.
References #
- R. T. Rockafellar, Convex Analysis, Princeton University Press, 1970, §9.
The sum map (x₁, …, xₘ) ↦ x₁ + ⋯ + xₘ of a finite product, as a linear map: the m-ary
codiagonal.
Equations
- Tdaf.ConvexAnalysis.piSum = { toFun := fun (x : ι → E) => ∑ i : ι, x i, map_add' := ⋯, map_smul' := ⋯ }
Instances For
A sum of finitely many sets is the image of their product under the sum map: what turns a question about a finite sum into a question about a linear image.
The cancellation hypothesis for a family, transported to the product set.
Closedness of a finite sum: a finite sum of closed convex sets is closed as soon as the only way finitely many directions of recession can sum to zero is inside the lineality spaces.
Closure distributes over a finite sum: cl (C₁ + ⋯ + Cₘ) = cl C₁ + ⋯ + cl Cₘ.
The recession cone of a finite sum of closed convex sets:
0⁺(cl C₁ + ⋯ + cl Cₘ) = 0⁺(cl C₁) + ⋯ + 0⁺(cl Cₘ).