Remark after Theorem 2, p. 910 — f(u) = ½u + ½√(u² + y′Σy) is nondecreasing in u
ProvedTractableDRO.MeanCov.remark_f_monotonedistributionally-robust-optimizationmoment-boundsp2o-batch-pfp2bp2o-gran-per-chapterp2o-plan-paperp2o-v1
Let be a symmetric positive semidefinite matrix and . Then the function
is nondecreasing on all of .
In the paper this observation is what turns the inner supremum over the mean set in (22) into a single linear constraint, giving the conic reformulation (23) of .
Formalization Note Positive semidefiniteness of (a covariance matrix in the paper) makes ; it is required, because with a negative constant under the root Lean's Real.sqrt returns for negative arguments and the function is not monotone.
Preamble
import Mathlib import Definitions.Def_TractableDRO_MeanCov_Model open Matrix
Formal statement
namespace TractableDRO.MeanCov
theorem remark_f_monotone {N : ℕ} (Sig : Matrix (Fin N) (Fin N) ℝ) (hSig : Sig.PosSemidef)
(y : Fin N → ℝ) :
Monotone (fun u : ℝ => u / 2 + Real.sqrt (u ^ 2 + y ⬝ᵥ Sig *ᵥ y) / 2) := by sorry
end TractableDRO.MeanCov
Source
Goh & Sim, Distributionally Robust Optimization and Its Tractable Approximations, Oper. Res. 58(4), 2010, p. 910, Remark after Theorem 2
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.