OAI/NumberTheory/JointDickman/Probability/ResidueCentering.lean

GitHub source file

17 declarations, 0 contain sorry, 0 declared as axioms

Centering finite channels in the residue coordinate

All residues above the same base cell carry the same atom weight. Thus subtracting the residue average of the transition subtracts exactly the residue average of the channel. The corresponding two-split kernel is centered in both coordinates. A residue-independent approximation cancels, at a loss of at most four in its uniform error.

  1. def OAI.JointDickman.finiteResidueAverage line 19
  2. theorem OAI.JointDickman.finiteResidueAverage_add line 22
  3. theorem OAI.JointDickman.finiteResidueAverage_sub line 27
  4. theorem OAI.JointDickman.finiteResidueAverage_mul_const line 32
  5. theorem OAI.JointDickman.finiteResidueAverage_const_mul line 37
  6. theorem OAI.JointDickman.finiteResidueAverage_div_const line 42
  7. theorem OAI.JointDickman.finiteResidueAverage_sum line 47
  8. theorem OAI.JointDickman.finiteResidueAverage_const line 52
  9. theorem OAI.JointDickman.finiteResidueAverage_abs_le line 58
  10. def OAI.JointDickman.residueCenteredTransition line 73

    Source documentation

    Subtract the uniform residue average within each base cell.

    noncomputable def residueCenteredTransition {Ω D R : Type*} [Fintype R]
        (p : Ω → D × R → ℝ) (s : Ω) (a : D × R) : ℝ :=
    …
  11. theorem OAI.JointDickman.residueCenteredTransition_average line 77
  12. theorem OAI.JointDickman.finiteChannel_residueCentered line 87

    Source documentation

    Equal cell weights make transition centering exactly channel centering. The identity also holds at zero weights, under the usual totalized division.

    theorem finiteChannel_residueCentered {Ω D R : Type*} [Fintype Ω] [Fintype D] [Fintype R]
        (w : Ω → ℝ) (μ : D → ℝ) (p : Ω → D × R → ℝ) (f : Ω → ℝ) (d : D) (r : R) :
        finiteChannel w (fun a : D × R => μ a.1) (residueCenteredTransition p) f (d, r) =
          finiteChannel w (fun a : D × R => μ a.1) p f (d, r) -
            finiteResidueAverage (fun t => finiteChannel w (fun a : D × R => μ a.1) p f (d, t)) := by
    …

    Mentions: OAI.JointDickman.finiteChannel, OAI.JointDickman.residueCenteredTransition, OAI.JointDickman.finiteResidueAverage

  13. def OAI.JointDickman.residueCenteredKernel line 102

    Source documentation

    Center a kernel in both of its residue coordinates.

    noncomputable def residueCenteredKernel {D R : Type*} [Fintype R]
        (K : D × R → D × R → ℝ) (a b : D × R) : ℝ :=
    …
  14. theorem OAI.JointDickman.finiteTwoSplitKernel_residueCentered line 110

    Source documentation

    The kernel of the centered channel is the actual four-term centered two-split kernel. No nonzero-weight assumption is needed for this identity.

    theorem finiteTwoSplitKernel_residueCentered {Ω D R : Type*}
        [Fintype Ω] [Fintype R] (w : Ω → ℝ) (μ : D → ℝ) (p : Ω → D × R → ℝ)
        (d e : D) (r s : R) :
        finiteTwoSplitKernel w (fun a : D × R => μ a.1) (residueCenteredTransition p) (d, r) (e, s) =
          residueCenteredKernel (finiteTwoSplitKernel w (fun a : D × R => μ a.1) p) (d, r) (e, s) := by
    …

    Mentions: OAI.JointDickman.finiteTwoSplitKernel, OAI.JointDickman.residueCenteredTransition, OAI.JointDickman.residueCenteredKernel

  15. theorem OAI.JointDickman.residueCenteredKernel_sub_base line 128
  16. theorem OAI.JointDickman.residueCenteredKernel_abs_le line 138

    Source documentation

    A residue-independent approximating kernel disappears after centering; the four remaining averages each contribute at most the uniform error.

    theorem residueCenteredKernel_abs_le {D R : Type*} [Fintype R] [Nonempty R]
        (K : D × R → D × R → ℝ) (G : D → D → ℝ) {ε : ℝ}
        (hK : ∀ d e r s, |K (d, r) (e, s) - G d e| ≤ ε) (d e : D) (r s : R) :
        |residueCenteredKernel K (d, r) (e, s)| ≤ 4 * ε := by
    …

    Mentions: OAI.JointDickman.residueCenteredKernel

  17. theorem OAI.JointDickman.finiteTwoSplitKernel_centered_abs_le line 158

    Source documentation

    Application of the uniform centering bound to the actual centered two-split kernel of the transition.

    theorem finiteTwoSplitKernel_centered_abs_le {Ω D R : Type*}
        [Fintype Ω] [Fintype R] [Nonempty R]
        (w : Ω → ℝ) (μ : D → ℝ) (p : Ω → D × R → ℝ) (G : D → D → ℝ) {ε : ℝ}
        (hK : ∀ d e r s,
          |finiteTwoSplitKernel w (fun a : D × R => μ a.1) p (d, r) (e, s) - G d e| ≤ ε)
        (d e : D) (r s : R) :
        |finiteTwoSplitKernel w (fun a : D × R => μ a.1) (residueCenteredTransition p)
          (d, r) (e, s)| ≤ 4 * ε := by
    …

    Mentions: OAI.JointDickman.finiteTwoSplitKernel, OAI.JointDickman.residueCenteredTransition