Lemma 4.4, p. 10 — for monotone M the embedding operator Q = F ∪ I is maximal monotone
ProvedHomogLCP.Embed.lemma_4_4Let be monotone, , and . Let be the Andersen–Ye operator (4.1) on and the infeasibility operator (4.7) with domain . Then the operator
on is maximal monotone: it is monotone, and it is not strictly contained in any other monotone operator on .
Since the normal cone operator is also maximal monotone, this makes the homogeneous embedding of a sum of two maximal monotone operators, to which Douglas–Rachford splitting applies with guaranteed convergence (Algorithm 5.1).
Formalization Note " is monotone" is the standing assumption of §§3–4 and is the explicit hypothesis (M + Mᵀ).PosSemidef; without it Lemma 4.1 and the statement fail. The cone does not enter and is not a hypothesis. Maximality is the containment form over all monotone operators (graphs) on (Fin d → ℝ) × ℝ. The case is allowed and the statement is true there.
import Mathlib import Definitions.Def_HomogLCP_Embed_MonotoneOp import Definitions.Def_HomogLCP_Embed_Setting
namespace HomogLCP.Embed
open Matrix
/-- Lemma 4.4, p. 10: for monotone `M`, the operator `𝒬 = ℱ ∪ ℐ` is maximal monotone. -/
theorem lemma_4_4 {d : ℕ} (M : Matrix (Fin d) (Fin d) ℝ) (q : Fin d → ℝ)
(hM : (M + Mᵀ).PosSemidef) :
IsMaximalMonotoneOp (embQ M q) := by sorry
end HomogLCP.Embed
Confirmed by the mission captain (proposal self-audit).
Confirmed by the moderator at approval.