AutoSamplingTheory.TechnicalLemmas.Analysis.PairingRockafellarSubgradient
9 named declarations scanned from AutoSamplingTheory/TechnicalLemmas/Analysis/PairingRockafellarSubgradient.lean.
Declarations
theorem AutoSamplingTheory.TechnicalLemmas.Analysis.PairingRockafellarSubgradient.chainValue_concat_singleton Partial Not mapped
- Appending one pair point to a finite chain adds exactly the final affine increment.
theorem chainValue_concat_singleton
(l : List (E × E)) (p : E × E) (z : E) :
chainValue (l ++ [p]) z =
chainValue l p.1 + inner ℝ p.2 (z - p.1) := by
simpa [chainValue] using chainValue_append_cons l p [] z
/-- A chain value at `x` can be extended by any relation point `(x,y)`. -/
AutoSamplingTheory/TechnicalLemmas/Analysis/PairingRockafellarSubgradient.lean:39published source at 0e31a3cda412
theorem AutoSamplingTheory.TechnicalLemmas.Analysis.PairingRockafellarSubgradient.add_inner_mem_rockafellarValueSet Partial Not mapped
- A chain value at `x` can be extended by any relation point `(x,y)`.
theorem add_inner_mem_rockafellarValueSet
{base : E × E} {Gamma : Set (E × E)}
{x y z : E} {r : ℝ}
(hxy : (x, y) ∈ Gamma)
(hr : r ∈ rockafellarValueSet base Gamma x) :
r + inner ℝ y (z - x) ∈ rockafellarValueSet base Gamma z := by
rcases hr with ⟨l, hlne, hhead, hforall, rfl⟩
refine ⟨l ++ [(x, y)], by simp [hlne], ?_, ?_, ?_⟩
· rw [List.head?_append_of_ne_nil l hlne]
exact hhead
· rw [List.forall_iff_forall_mem] at hforall ⊢
intro p hp
rw [List.mem_append] at hp
rcases hp with hp | hp
· exact hforall p hp
· simp at hp
simpa [hp] using hxy
· simpa [chainValue_concat_singleton]
/-- Closed-chain nonpositivity supplies an explicit finite upper bound for every
rooted chain value at the source coordinate of a relation point. -/
AutoSamplingTheory/TechnicalLemmas/Analysis/PairingRockafellarSubgradient.lean:46published source at 0e31a3cda412
theorem AutoSamplingTheory.TechnicalLemmas.Analysis.PairingRockafellarSubgradient.rockafellarValueSet_le_inner_of_mem Partial Not mapped
- Closed-chain nonpositivity supplies an explicit finite upper bound for every rooted chain value at the source coordinate of a relation point.
theorem rockafellarValueSet_le_inner_of_mem
{base : E × E} {Gamma : Set (E × E)}
(hclosed : PairingClosedChainMonotone Gamma)
{x y : E} (hxy : (x, y) ∈ Gamma)
{r : ℝ} (hr : r ∈ rockafellarValueSet base Gamma x) :
r ≤ inner ℝ y (x - base.1) := by
rcases hr with ⟨l, hlne, hhead, hforall, rfl⟩
have hforallAppend :
List.Forall (fun p => p ∈ Gamma) (l ++ [(x, y)]) := by
rw [List.forall_iff_forall_mem] at hforall ⊢
intro p hp
rw [List.mem_append] at hp
rcases hp with hp | hp
· exact hforall p hp
· simp at hp
simpa [hp] using hxy
have hclosedAppend :
chainValue (l ++ [(x, y)]) base.1 ≤ 0 :=
hclosed (l := l ++ [(x, y)]) (base := base)
(by simp [hlne])
(by
rw [List.head?_append_of_ne_nil l hlne]
exact hhead)
hforallAppend
rw [chainValue_concat_singleton] at hclosedAppend
have hreverse :
inner ℝ y (base.1 - x) = -inner ℝ y (x - base.1) := by
rw [show base.1 - x = -(x - base.1) by abel, inner_neg_right]
rw [hreverse] at hclosedAppend
linarith
/-- The extended value set is always bounded above by `⊤`. -/
AutoSamplingTheory/TechnicalLemmas/Analysis/PairingRockafellarSubgradient.lean:67published source at 0e31a3cda412
theorem AutoSamplingTheory.TechnicalLemmas.Analysis.PairingRockafellarSubgradient.bddAbove_properRockafellarValueSet Partial Not mapped
- The extended value set is always bounded above by `⊤`.
theorem bddAbove_properRockafellarValueSet
(base : E × E) (Gamma : Set (E × E)) (x : E) :
BddAbove (properRockafellarValueSet base Gamma x) :=
⟨⊤, fun _ _ => le_top⟩
/-- Every explicit rooted-chain value lies below the extended Rockafellar
supremum. -/
AutoSamplingTheory/TechnicalLemmas/Analysis/PairingRockafellarSubgradient.lean:99published source at 0e31a3cda412
theorem AutoSamplingTheory.TechnicalLemmas.Analysis.PairingRockafellarSubgradient.coe_le_properRockafellarPotential_of_mem Partial Not mapped
- Every explicit rooted-chain value lies below the extended Rockafellar supremum.
theorem coe_le_properRockafellarPotential_of_mem
{base : E × E} {Gamma : Set (E × E)} {x : E} {r : ℝ}
(hr : r ∈ rockafellarValueSet base Gamma x) :
((r : ℝ) : WithTop ℝ) ≤ properRockafellarPotential base Gamma x := by
rw [properRockafellarPotential]
exact le_csSup (bddAbove_properRockafellarValueSet base Gamma x)
(Set.mem_image_of_mem ((↑) : ℝ → WithTop ℝ) hr)
/-- At every source coordinate appearing in `Gamma`, the proper Rockafellar
potential is finite. -/
AutoSamplingTheory/TechnicalLemmas/Analysis/PairingRockafellarSubgradient.lean:106published source at 0e31a3cda412
theorem AutoSamplingTheory.TechnicalLemmas.Analysis.PairingRockafellarSubgradient.properRockafellarPotential_lt_top_of_mem Partial Not mapped
- At every source coordinate appearing in `Gamma`, the proper Rockafellar potential is finite.
theorem properRockafellarPotential_lt_top_of_mem
{base : E × E} {Gamma : Set (E × E)}
(hbase : base ∈ Gamma)
(hclosed : PairingClosedChainMonotone Gamma)
{x y : E} (hxy : (x, y) ∈ Gamma) :
properRockafellarPotential base Gamma x < ⊤ := by
have hupper :
properRockafellarPotential base Gamma x ≤
((inner ℝ y (x - base.1) : ℝ) : WithTop ℝ) := by
rw [properRockafellarPotential]
refine csSup_le (properRockafellarValueSet_nonempty hbase x) ?_
intro u hu
rcases hu with ⟨r, hr, rfl⟩
exact WithTop.coe_le_coe.mpr
(rockafellarValueSet_le_inner_of_mem hclosed hxy hr)
exact lt_of_le_of_lt hupper (WithTop.coe_lt_top _)
/-- Extended-real supporting-vector relation. The existential real value makes
finiteness at the contact point part of the proposition instead of a hidden
precondition. -/
AutoSamplingTheory/TechnicalLemmas/Analysis/PairingRockafellarSubgradient.lean:116published source at 0e31a3cda412
def AutoSamplingTheory.TechnicalLemmas.Analysis.PairingRockafellarSubgradient.ProperSupportsAt Partial Not mapped
- Extended-real supporting-vector relation. The existential real value makes finiteness at the contact point part of the proposition instead of a hidden precondition.
def ProperSupportsAt
(Phi : E → WithTop ℝ) (x y : E) : Prop :=
∃ rx : ℝ,
Phi x = ((rx : ℝ) : WithTop ℝ) ∧
∀ z : E,
(((rx + inner ℝ y (z - x) : ℝ) : ℝ) : WithTop ℝ) ≤ Phi z
/-- Every point of a closed-chain-monotone relation supports the proper
Rockafellar potential at its source coordinate. -/
AutoSamplingTheory/TechnicalLemmas/Analysis/PairingRockafellarSubgradient.lean:136published source at 0e31a3cda412
theorem AutoSamplingTheory.TechnicalLemmas.Analysis.PairingRockafellarSubgradient.properSupportsAt_of_mem Partial Not mapped
- Every point of a closed-chain-monotone relation supports the proper Rockafellar potential at its source coordinate.
theorem properSupportsAt_of_mem
{base : E × E} {Gamma : Set (E × E)}
(hbase : base ∈ Gamma)
(hclosed : PairingClosedChainMonotone Gamma)
{x y : E} (hxy : (x, y) ∈ Gamma) :
ProperSupportsAt (properRockafellarPotential base Gamma) x y := by
have hxfin := properRockafellarPotential_lt_top_of_mem hbase hclosed hxy
have hxne : properRockafellarPotential base Gamma x ≠ ⊤ := ne_of_lt hxfin
let rx : ℝ :=
(properRockafellarPotential base Gamma x).untop hxne
have hxcoe :
properRockafellarPotential base Gamma x = ((rx : ℝ) : WithTop ℝ) := by
exact (WithTop.coe_untop _ hxne).symm
refine ⟨rx, hxcoe, ?_⟩
intro z
by_cases hz : properRockafellarPotential base Gamma z = ⊤
· rw [hz]
exact le_top
· let rz : ℝ :=
(properRockafellarPotential base Gamma z).untop hz
have hzcoe :
properRockafellarPotential base Gamma z = ((rz : ℝ) : WithTop ℝ) := by
exact (WithTop.coe_untop _ hz).symm
have hupper :
properRockafellarPotential base Gamma x ≤
(((rz - inner ℝ y (z - x) : ℝ) : ℝ) : WithTop ℝ) := by
rw [properRockafellarPotential]
refine csSup_le (properRockafellarValueSet_nonempty hbase x) ?_
intro u hu
rcases hu with ⟨r, hr, rfl⟩
have hrext :
r + inner ℝ y (z - x) ∈ rockafellarValueSet base Gamma z :=
add_inner_mem_rockafellarValueSet hxy hr
have hle := coe_le_properRockafellarPotential_of_mem hrext
rw [hzcoe] at hle
exact WithTop.coe_le_coe.mpr (by
have hreal := WithTop.coe_le_coe.mp hle
linarith)
rw [hxcoe] at hupper
have hupperReal := WithTop.coe_le_coe.mp hupper
rw [hzcoe]
exact WithTop.coe_le_coe.mpr (by linarith)
/-- Direct bridge from the transport-produced distinct-cycle condition. -/
AutoSamplingTheory/TechnicalLemmas/Analysis/PairingRockafellarSubgradient.lean:145published source at 0e31a3cda412
theorem AutoSamplingTheory.TechnicalLemmas.Analysis.PairingRockafellarSubgradient.properSupportsAt_of_mem_of_distinct Partial Not mapped
- Direct bridge from the transport-produced distinct-cycle condition.
theorem properSupportsAt_of_mem_of_distinct
{base : E × E} {Gamma : Set (E × E)}
(hbase : base ∈ Gamma)
(hmono : PairingDistinctCycleMonotone Gamma)
{x y : E} (hxy : (x, y) ∈ Gamma) :
ProperSupportsAt (properRockafellarPotential base Gamma) x y :=
properSupportsAt_of_mem hbase
(pairingClosedChainMonotone_of_distinct hmono) hxy
end
end PairingRockafellarSubgradient
end Analysis
end TechnicalLemmas
end AutoSamplingTheory
AutoSamplingTheory/TechnicalLemmas/Analysis/PairingRockafellarSubgradient.lean:189published source at 0e31a3cda412