RDependent #
Ported from scottnarmstrong/Superdiffusion at 7f20476385d4af418f110303085ad0740c9eba46. Module imports, support namespace and local option directives adapted for Lean Pool.
r-dependence of a ℤ-indexed family, and the ZMod (r+1) colouring #
ABK26, Proposition p.concentration, specialised to a ℤ-indexed sequence.
"Since the sequence
{X_j}is 2-dependent, we may apply Propositionp.concentration…"
Why r and not 2 #
The manuscript's "2-dependent" is dimension-restricted, while the reusable
predicate here records dependence at an arbitrary separation radius r.
The colouring #
Splitting a window [n, m] ⊆ ℤ into the r+1 residue classes mod r+1 makes
two distinct indices of one class differ by a nonzero multiple of r+1, hence be
≥ r-separated, hence make the class mutually independent. That separation
property is le_abs_sub_of_intCast_zmod_eq; the palette size is
card_zmod_succ.
Main results #
Algsuperdiff.Probability.RDependent.comp,.monoAlgsuperdiff.Probability.le_abs_sub_of_intCast_zmod_eqAlgsuperdiff.Probability.card_zmod_succ
References #
- ABK26, Proposition
p.concentration. - ABK26,
l.minimal.scale.sep, Step 1 (for the separation claim).
The r-dependence hypothesis #
A family X : ℤ → Ω → ℝ is r-dependent: any finite set of pairwise ≥ r-separated indices is mutually independent. This is the hypothesis of ABK26
p.concentration specialised to a ℤ-indexed sequence, and the hypothesis
under which Step 1 of l.minimal.scale.sep invokes it.
Equations
- Algsuperdiff.Probability.RDependent P X r = ∀ (s : Finset ℤ), (∀ i ∈ s, ∀ j ∈ s, i ≠ j → ↑r ≤ |i - j|) → ProbabilityTheory.iIndepFun (fun (i : ↥s) => X ↑i) P
Instances For
r-dependence is stable under composing each coordinate with a measurable map
(used to pass from X to the centred family X − E X).
The colouring's separation property. Two distinct integers with the same
residue mod r+1 differ by a nonzero multiple of r+1, hence by at least
r+1. (For r = 2 this is "same residue mod 3 ⟹ |i − j| > 1".)
The ZMod (r+1) palette has exactly r+1 colours.