Moreau decomposition

TheoremVariational characterization of projection onto a convex cone#✓fully proved

Variational characterization of the projection onto a cone with complete carrier: p is the projection of x iff the residual pairs nonpositively with the cone and is orthogonal to p. The ambient space need not be complete.

theorem convexProjection_eq_iff (C : PointedCone ℝ E) (hC : IsComplete (C : Set E))
    (x p : E) (hp : p ∈ C) :
    convexProjection (C : Set E) x = p ↔
      (∀ y ∈ C, ⟪x - p, y⟫_ℝ ≤ 0) ∧ ⟪x - p, p⟫_ℝ = 0
Uses 3 · Used by 3

TheoremCone projection residual pairs nonpositively with the cone#✓fully proved

The residual of the cone projection pairs nonpositively with every point of the cone.

theorem convexProjection_inner_le_zero (C : PointedCone ℝ E)
    (hC : IsComplete (C : Set E)) (x y : E) (hy : y ∈ C) :
    ⟪x - convexProjection (C : Set E) x, y⟫_ℝ ≤ 0
Uses 3 · Used by 1

TheoremCone projection residual is orthogonal to the projection#✓fully proved

The residual of the cone projection is orthogonal to the projection.

theorem convexProjection_inner_self (C : PointedCone ℝ E) (hC : IsComplete (C : Set E))
    (x : E) :
    ⟪x - convexProjection (C : Set E) x, convexProjection (C : Set E) x⟫_ℝ = 0
Uses 3 · Used by 2

TheoremVariational characterization determines the cone projection#✓fully proved

A point satisfying the cone variational characterization is the projection.

theorem convexProjection_eq_of_forall_inner_le_zero (C : PointedCone ℝ E)
    (hC : IsComplete (C : Set E)) {x p : E} (hp : p ∈ C)
    (h1 : ∀ y ∈ C, ⟪x - p, y⟫_ℝ ≤ 0) (h2 : ⟪x - p, p⟫_ℝ = 0) :
    convexProjection (C : Set E) x = p
Uses 2 · Used by 1

DefinitionNegative polar cone#✓fully proved

The negative polar is the standard dual cone of the negated set. Continuity makes its carrier closed without requiring completeness of the ambient space.

def polarCone (K : Set E) : ProperCone ℝ E where
  toSubmodule := PointedCone.dual (innerₗ E) (-K)
  isClosed' := PointedCone.isClosed_dual fun x => (innerSL ℝ x).continuous
Used by 6

TheoremMembership in the negative polar cone#✓fully proved

Membership in the negative polar uses the nonpositive inner-product convention.

theorem mem_polarCone {K : Set E} {y : E} :
    y ∈ polarCone K ↔ ∀ x ∈ K, ⟪x, y⟫_ℝ ≤ 0
Uses 1 · Used by 3

TheoremCone projection residual lies in the negative polar#✓fully proved

The residual of a cone projection lies in its negative polar.

theorem sub_convexProjection_mem_polarCone (C : PointedCone ℝ E)
    (hC : IsComplete (C : Set E)) (x : E) :
    x - convexProjection (C : Set E) x ∈ polarCone (C : Set E)
Uses 4 · Used by 1

TheoremNegative polar agrees with the inner dual of the negated cone#✓fully proved

In a complete ambient space, the negative polar agrees with Mathlib's inner dual of the negated carrier.

theorem polarCone_eq_innerDual_neg (K : Set E) :
    polarCone K = ProperCone.innerDual (-K)
Uses 1
Statement usesNegative polar cone

TheoremMoreau decomposition, polar form#✓fully proved

The cone projection's residual is the projection onto its negative polar. This accepts a Mathlib proper cone directly.

theorem convexProjection_polarCone (C : ProperCone ℝ E) (x : E) :
    convexProjection (polarCone (C : Set E)) x = x - convexProjection (C : Set E) x
Uses 7 · Used by 1

TheoremMoreau decomposition#✓fully proved

A point is the sum of its projections onto a proper cone and its negative polar, and the projections are orthogonal.

theorem moreau_decomposition (C : ProperCone ℝ E) (x : E) :
    x = convexProjection (C : Set E) x + convexProjection (polarCone (C : Set E)) x ∧
      ⟪convexProjection (C : Set E) x, convexProjection (polarCone (C : Set E)) x⟫_ℝ = 0
Uses 4

TheoremBipolar theorem for closed convex cones#✓fully proved

The negative bipolar of a proper cone is itself, by Mathlib's inner-dual bipolar theorem.

theorem polarCone_polarCone (C : ProperCone ℝ E) :
    polarCone (polarCone (C : Set E) : Set E) = C
Uses 2
11 fully proved

The dependency explorer needs JavaScript and its graph data.

Download the complete graph data.