Figure 2.1, pp.98–99 — is not convex
ProvedKallMayerSLP.ProbConstr.figure_2_1_examplechance-constraintsp2o-gran-per-chapterp2o-plan-bookp2o-v1probability-constraintsstochastic-programming
Let have a finite discrete distribution with three realizations of probabilities , , , and let the data be such that the realized constraint sets in are
Then at the level
and is not convex.
The example shows that can fail to be convex for a finite discrete distribution at a level below the thresholds of Theorems 2.3 and 2.4.
Differences from the source:
- reading: the sets K_1, K_2, K_3 and the probabilities, with no data T, h, ξ̂ given → every tableau with p = (1/4, 1/2, 1/4) and every data T, h (any r, any number m of rows) whose realized sets are the three printed sets; reason: the book specifies only the K_j; B(α) depends on the data only through the K_j and p
- encoding: indices 1, 2, 3 and coordinates x_1, x_2 → Fin 3 indices 0, 1, 2 and coordinates x 0, x 1; reason: renaming only
Preamble
import Mathlib import Definitions.Def_KallMayerSLP_ProbConstr_realizationSet import Definitions.Def_KallMayerSLP_ProbConstr_feasibleSet import Definitions.Def_KallMayerSLP_ProbConstr_RealizationTableau_measure
Formal statement
namespace KallMayerSLP.ProbConstr
/-- The example of Figure 2.1 (from Kall [134]): three realizations with
`K₁ = {x ∈ ℝ² | x₁ − x₂ ≥ −2, x₂ ≥ 3}`, `K₂ = {x ∈ ℝ² | x₁ − x₂ ≥ 0, 2x₁ + 3x₂ ≤ 25}`,
`K₃ = {x ∈ ℝ² | x₁ + x₂ ≤ 8, −x₁ + 3x₂ ≥ 0}` and probabilities `p₁ = 1/4`, `p₂ = 1/2`,
`p₃ = 1/4`, at level `α = 3/4`: the feasible domain is non-convex and
`B(α) = [K₁ ∩ K₂] ∪ [K₂ ∩ K₃]`.
Kall & Mayer 2005, Chapter 2, §2.2, p.98 (PDF 108), and Figure 2.1, p.99 (PDF 109).
**Formalization Note.** Indices `0, 1, 2` stand for the book's `1, 2, 3`, coordinates `x 0`,
`x 1` for `x₁`, `x₂`. The book gives the sets `K_j` and not the data; the theorem holds for every
tableau with these probabilities and every data `T`, `h` (any number `m` of rows, any `r`) whose
realized sets `K_j = {x | T(ξ̂^j)x ≥ h(ξ̂^j)}` are the three printed sets. -/
theorem figure_2_1_example {r m : ℕ} (τ : RealizationTableau 3 r)
(hp : τ.p 0 = 1 / 4 ∧ τ.p 1 = 1 / 2 ∧ τ.p 2 = 1 / 4)
(T : (Fin r → ℝ) → Matrix (Fin m) (Fin 2) ℝ) (h : (Fin r → ℝ) → Fin m → ℝ)
(hK₁ : realizationSet T h (τ.xi 0) = {x | x 0 - x 1 ≥ -2 ∧ x 1 ≥ 3})
(hK₂ : realizationSet T h (τ.xi 1) = {x | x 0 - x 1 ≥ 0 ∧ 2 * x 0 + 3 * x 1 ≤ 25})
(hK₃ : realizationSet T h (τ.xi 2) = {x | x 0 + x 1 ≤ 8 ∧ -x 0 + 3 * x 1 ≥ 0}) :
feasibleSet τ.measure T h (3 / 4) =
(realizationSet T h (τ.xi 0) ∩ realizationSet T h (τ.xi 1)) ∪
(realizationSet T h (τ.xi 1) ∩ realizationSet T h (τ.xi 2)) ∧
¬ Convex ℝ (feasibleSet τ.measure T h (3 / 4)) := by sorry
end KallMayerSLP.ProbConstr
Source
Kall & Mayer, Stochastic Linear Programming: Models, Theory, and Computation (Kluwer/Springer 2005), Chapter 2, Chapter 2, §2.2, p.98 (PDF 108), example from Kall [134], and Figure 2.1, p.99 (PDF 109)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.