OAI/Probability/DilutedSpin/ShapeProjectionEnergy.lean

GitHub source file

8 declarations, 0 contain sorry, 0 declared as axioms

  1. lemma OAI.DilutedSpinGlass.PrescribedTree.branch_leaf_triple line 12

    Source documentation

    Pulling three independent copies of one sampled branch back to any one leaf keeps exactly the three independent path law.

    lemma branch_leaf_triple (C : PrescribedTree n) (T : KernelTower Ω (n+1))
        (a : C.Leaf) (F : FinitePath Ω (n+1) → FinitePath Ω (n+1) →
          FinitePath Ω (n+1) → ℝ) :
        (T.1.bind (fun z => C.sampleLaw (T.2 z))).expect (fun x =>
          (T.1.bind (fun z => C.sampleLaw (T.2 z))).expect (fun y =>
            (T.1.bind (fun z => C.sampleLaw (T.2 z))).expect (fun w =>
              F (x.1,C.pathAt a x.2) (y.1,C.pathAt a y.2) (w.1,C.pathAt a w.2)))) =
        KernelTower.tripleExpectAt (n+1) T 0 F := by
    …

    Mentions: OAI.DilutedSpinGlass.PrescribedTree, OAI.DilutedSpinGlass.KernelTower, OAI.DilutedSpinGlass.FinitePath, OAI.DilutedSpinGlass.KernelTower.tripleExpectAt

  2. def OAI.DilutedSpinGlass.PrescribedTree.branchProjectionSq line 42

    Source documentation

    The error is measured in the genuine two-copy branch law, in the same coupling as the physical spin products, not in a surrogate law.

    noncomputable def branchProjectionSq (C : PrescribedTree n) (T : KernelTower Ω (n+1))
        (a : C.Leaf) (f Y : FinitePath Ω (n+1) → Fin N → ℝ) : ℝ :=
    …

    Mentions: OAI.DilutedSpinGlass.PrescribedTree, OAI.DilutedSpinGlass.KernelTower, OAI.DilutedSpinGlass.FinitePath

  3. lemma OAI.DilutedSpinGlass.PrescribedTree.branch_energy_projection_transfer line 48
  4. def OAI.DilutedSpinGlass.PrescribedTree.shapeEnergyAt line 63

    Source documentation

    Conditional covariance of the actual shape with d preceding unary transitions, averaged over their genuine kernel prefix.

    noncomputable def shapeEnergyAt (C : PrescribedTree n) : (d : ℕ) →
        KernelTower Ω (n+1+d) → (FinitePath Ω (n+1+d) → Fin N → ℝ) → ℝ
      | 0, T, f => childEnergy C T f
      | d+1, T, f => T.1.expect (fun z => shapeEnergyAt C d (T.2 z) (fun y => f (z,y)))

    Mentions: OAI.DilutedSpinGlass.PrescribedTree, OAI.DilutedSpinGlass.KernelTower, OAI.DilutedSpinGlass.FinitePath, OAI.DilutedSpinGlass.PrescribedTree.childEnergy

  5. def OAI.DilutedSpinGlass.PrescribedTree.shapeProjectionSqAt line 68
  6. theorem OAI.DilutedSpinGlass.PrescribedTree.shape_energy_projection_transfer line 77

    Source documentation

    Three-copy transfer at an arbitrary prefix above a complete literal multileaf shape. The shape's actual covariance is on the left.

    theorem shape_energy_projection_transfer (C : PrescribedTree n) (a : C.Leaf)
        (d : ℕ) (T : KernelTower Ω (n+1+d))
        (f Y : FinitePath Ω (n+1+d) → Fin N → ℝ) :
        shapeEnergyAt C d T f ≤ 2*KernelTower.halfTripleDifferenceAt (n+1+d) T d Y +
          4*shapeProjectionSqAt C a d T f Y := by
    …

    Mentions: OAI.DilutedSpinGlass.PrescribedTree, OAI.DilutedSpinGlass.KernelTower, OAI.DilutedSpinGlass.FinitePath, OAI.DilutedSpinGlass.PrescribedTree.shapeEnergyAt, OAI.DilutedSpinGlass.KernelTower.halfTripleDifferenceAt, OAI.DilutedSpinGlass.PrescribedTree.shapeProjectionSqAt

  7. lemma OAI.DilutedSpinGlass.PrescribedTree.shapeEnergyAt_nonneg line 92
  8. lemma OAI.DilutedSpinGlass.PrescribedTree.shapeProjectionSqAt_nonneg line 98