§3.2, proof of Corollary 3.5, p. 485 — y* ∈ ∂h(x*) ∩ ∂g(x*) gives x* ∈ ∂g*(y*) ∩ ∂h*(y*)
ProvedTaoAnDCA.LocalOpt.mem_subdiff_conj_of_mem_subdiffconjugatedc-programmingp2o-batch-pfp2ap2o-gran-per-chapterp2o-plan-paperp2o-v1subdifferential
Let satisfy the standing inclusions and . If and , then
In the proof of Corollary 3.5 this inverts the subgradient relation: after Theorem 3.2(i) gives , the point is a common subgradient of the two dual components at , i.e. is a critical point of .
Formalization Note. Membership uses the published IsSubgradient, which also asserts ; the conjugate is the EReal supremum CondatPD.FinDim.conj.
Preamble
import Mathlib import Definitions.Def_TaoAnDCA_LocalOpt_Setting open Filter Topology
Formal statement
namespace TaoAnDCA.LocalOpt
/-- §3.2, proof of Corollary 3.5, p. 485: if `y* ∈ ∂h(x*)` and `y* ∈ ∂g(x*)`, then
`x* ∈ ∂g*(y*) ∩ ∂h*(y*)`. -/
theorem mem_subdiff_conj_of_mem_subdiff {n : ℕ} (g h : EuclideanSpace ℝ (Fin n) → EReal)
(hgh : TaoAnDCA.GlobalOpt.DCStanding g h) (xs ys : EuclideanSpace ℝ (Fin n))
(hy : ys ∈ TaoAnDCA.GlobalOpt.subdiff h xs) (hyg : ys ∈ TaoAnDCA.GlobalOpt.subdiff g xs) :
xs ∈ TaoAnDCA.GlobalOpt.subdiff (CondatPD.FinDim.conj g) ys ∩ TaoAnDCA.GlobalOpt.subdiff (CondatPD.FinDim.conj h) ys := 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. 485, §3.2, proof of Corollary 3.5, second sentence
Human review
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.