Proposition 3.1 — each (P_s) attains its minimum, α̂ = min over 𝒞 of F is attained, equals E{α_s}, and bounds min (P) below
ProvedProgHedging.Convex.proposition_3_1p2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1progressive-hedgingstochastic-programming
Under the standing assumptions, each scenario subproblem — minimize over — has an optimal solution, hence a finite optimal value . Moreover the value
is attained, and for any choice of optimal solutions of ,
Finally , the optimal value of (P) (which is if (P) is infeasible).
This identifies the trivial lower bound obtained by solving the scenarios separately, the starting point suggested for the algorithm.
Formalization Note. is the EReal infimum of over ; its attainment is part of the statement. is written as for an arbitrary family of minimizers.
Preamble
import Mathlib import Definitions.Def_ProgHedging_Convex_Problem import Definitions.Def_ProgHedging_Convex_Duality open scoped RealInnerProductSpace Pointwise Topology open Filter
Formal statement
namespace ProgHedging.Convex
/-- Proposition 3.1, pp. 10–11. Each scenario subproblem `(P_s)` attains its minimum; the value
`α̂ = min_{X ∈ 𝒞} F(X)` is attained and equals `E{α_s}` with `α_s = min (P_s)`; and `α̂ ≤ min (P)`. -/
theorem proposition_3_1 {S : Type*} [Fintype S] {n T : ℕ} (pr : Problem S n T) :
(∀ s, ∃ x ∈ pr.C s, ∀ z ∈ pr.C s, pr.f s x ≤ pr.f s z) ∧
(∃ X ∈ pr.adm, ((pr.F X : ℝ) : EReal) = pr.alphaHat) ∧
(∀ xs : S → EuclideanSpace ℝ (Fin n),
(∀ s, xs s ∈ pr.C s ∧ ∀ z ∈ pr.C s, pr.f s (xs s) ≤ pr.f s z) →
pr.alphaHat = ((∑ s, pr.p s * pr.f s (xs s) : ℝ) : EReal)) ∧
pr.alphaHat ≤ pr.minP := by sorry
end ProgHedging.Convex
Source
Rockafellar and Wets, Scenarios and policy aggregation in optimization under uncertainty, IIASA Working Paper WP-87-119 (1987), pp. 10–11, Proposition 3.1, (3.2)–(3.3)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.