Resample #
Ported from scottnarmstrong/CoarseGraining at c7ddd76c08ade64fed1b8d2ca51be14dfee8deb4. Module imports, support namespace and local option directives adapted for Lean Pool.
One-core resampling stability #
The deterministic sensitivity estimate e.fixed.phase.sensitivity behind the
Efron–Stein step of Proposition 4.3. Resampling the coefficient field on a
single core coreBox ℓ σ k moves the fixed-phase observable
F_σ(b) = P · 𝐀(U; b_σ) P by at most a normalized energy over that core:
|F_σ(b) − F_σ(b')| ≤ 48 Θ (vol U)⁻¹ ∫_{coreBox ℓ σ k ∩ U} Z · 𝐀_σ(b) Z,
where Z is a minimizer for corridorField ℓ σ b and b, b' are two fields,
(1, Θ)-elliptic on U, agreeing off the core.
The proof specializes the corridor-independent core
abs_coarseObservable_sub_le_of_minimizer from PhaseComparison.Stability
to the two corridor-modified fields and the core intersection. The field
agreement and ellipticity inputs express the resampling-specific geometry;
the analytic minimizer assembly has one implementation.
Measurability of a core box #
Each core box is measurable (a finite product of closed intervals).
The one-core resampling specialization #
Part 3 (one-core resampling stability), minimizer form. For a fixed grid
phase σ and two fields b, b' that are (1, Θ)-elliptic on
U := cubeSet (originCube d m) and agree off the core coreBox ℓ σ k, the
fixed-phase observable moves by at most a normalized core-energy of the
minimizer Z for corridorField ℓ σ b.
Part 3 (one-core resampling stability). Existential-minimizer wrapper.