Theorem 3.2(i), p. 484 — a local minimizer x* of g − h satisfies ∂h(x*) ⊂ ∂g(x*)
ProvedTaoAnDCA.LocalOpt.theorem_3_2_idc-programminglocal-optimalityp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1subdifferential
Let satisfy the standing inclusions and , and let the d.c. difference be taken with the convention . If is a local minimizer of (that is, is finite and for all in a neighbourhood of ), then :
This is the classical necessary condition for local optimality in d.c. programming; it is the first step in transporting a local minimizer of the primal program to the dual one (Corollary 3.5).
Formalization Note. Local minimizer and are the definitions of TaoAnDCA.LocalOpt.Setting; the standing assumptions are the structure DCStanding g h.
Preamble
import Mathlib import Definitions.Def_TaoAnDCA_LocalOpt_Setting open Filter Topology
Formal statement
namespace TaoAnDCA.LocalOpt
/-- Theorem 3.2 (i), p. 484: a local minimizer `x*` of `g − h` lies in `𝒫_l`,
i.e. `∂h(x*) ⊂ ∂g(x*)`. -/
theorem theorem_3_2_i {n : ℕ} (g h : EuclideanSpace ℝ (Fin n) → EReal) (hgh : TaoAnDCA.GlobalOpt.DCStanding g h)
(xs : EuclideanSpace ℝ (Fin n)) (hloc : IsDCLocalMin g h xs) :
xs ∈ Pl g h := by sorry
end TaoAnDCA.LocalOpt
Source
Pham Dinh & Le Thi, A d.c. optimization algorithm for solving the trust-region subproblem, SIAM J. Optim. 8 (1998), p. 484, Theorem 3.2(i)
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.