Proof of Theorem 3.3, p. 591 ā E_{P_m(x)}[š(š¬(hĢ, x))] is at least the objective of (16)
ProvedMinimaxSLP.RhsExtremal.mixture_lower_boundgaussianmoment-problemp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1probability
Under Assumptions 2, 3 (at the constant ) and 5, and , let be a feasible point of (16) in which every block has or is zero, and let be the associated mixture of normal distributions. Then is -integrable and
Applied to an optimal solution of (16), the right-hand side is ; this is the lower half of the sandwich that proves Theorem 3.3.
Formalization Note The statement is made for every feasible point with zero-or-positive blocks, not only for an optimal one; the paper's last display, with "", is the specialization to an optimal point.
Preamble
import Mathlib import Definitions.Def_MomentDRO_Conf_Setting import Definitions.Def_MinimaxSLP_RhsExtremal_Model
Formal statement
namespace MinimaxSLP.RhsExtremal
open MeasureTheory Matrix
/-- Proof of Theorem 3.3, p. 591: for every feasible `s` of (16) whose blocks have
`v_k0^i > 0` or are zero, `š(š¬(Ā·, x))` is integrable under the mixture `P_m(x)` and
`E_{P_m(x)}[š(š¬(hĢ, x))] ℠Σ_{k,i} (α_k p_iā²v_k^i + v_k0^i (β_k ā α_k p_iā²T x))`, the objective
of (16) at `s`. -/
theorem mixture_lower_bound {r d n K N : ā} [NeZero K]
(W : Matrix (Fin r) (Fin d) ā) (T : Matrix (Fin r) (Fin n) ā) (q : Fin d ā ā)
(α β : Fin K ā ā) (ps : Fin N ā Fin r ā ā) (μ : Fin r ā ā) (Q : Matrix (Fin r) (Fin r) ā)
(hα : ā k, 0 ⤠α k)
(hA2 : ā z : Fin r ā ā, ā w : Fin d ā ā, 0 ⤠w ā§ W *ᵄ w = z)
(hA3 : ā Ļ : Fin r ā ā, Wįµ *ᵄ Ļ ā¤ q)
(hA5 : Set.range ps = Set.extremePoints ā {Ļ : Fin r ā ā | Wįµ *ᵄ Ļ ā¤ q})
(x : Fin n ā ā)
(s : Fin K ā Fin N ā (Matrix (Fin r) (Fin r) ā Ć (Fin r ā ā) Ć ā))
(hs : s ā feasible16 K N μ Q)
(hpos : ā k i, 0 < (s k i).2.2 ⨠s k i = 0) :
Integrable (fun h => MinimaxSLP.ObjSDP.disutility α β (MinimaxSLP.RhsSDP.Qval W T q h x)) (mixture s) ā§
obj16 T α β ps x s ⤠⫠h, MinimaxSLP.ObjSDP.disutility α β (MinimaxSLP.RhsSDP.Qval W T q h x) ā(mixture s) := by sorry
end MinimaxSLP.RhsExtremal
Source
Bertsimas, Doan, Natarajan & Teo, Models for Minimax Stochastic Linear Optimization Problems with Risk Aversion, Math. Oper. Res. 35(3), 2010, p. 591, proof of Theorem 3.3
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.