A drift bound retaining the far-mass loss
ProvedAnalyticCombinatorics.Ch8.Partitions.Erdos.abs_drift_le_softanalytic-combinatoricsdrift-coalescencelean4partitions-support
A one-step drift bound includes an explicit far-mass loss; summability/smallness of that tail over the required horizon remains an input.
The full literal source contract is:
theorem abs_drift_le_soft (b R : ℝ) (x : ι) (hbnn : 0 ≤ b)
(hKnn : ∀ z, 0 ≤ K x z) (hKsum : ∑ z, K x z = 1) (hR : ∀ z, |D z| ≤ R) :
(∑ z, K x z * |D z|) ≤ |D x| + b + 2 * R * farMass K D b x
Preamble
import Definitions.Def_AnalyticCombinatorics_Ch8_Partitions_SoftTailDrift
import Mathlib
set_option autoImplicit true
set_option relaxedAutoImplicit false
set_option maxSynthPendingDepth 3
/-!
# Soft-tail Tanaka drift bound
The banked `abs_drift_le` needs a *global* hard increment bound (`K x z ≠ 0 → |D z − D x| ≤ b`).
The Erdős predecessor kernel only has *soft* far tails: most mass moves by `≤ b`, but a small
`farMass` moves further (up to the range `2R`). This lemma charges that far mass to an additive
error: the one-step `|D|`-drift is `≤ b + 2R·farMass`, where
`farMass K D b x = ∑_z K x z · 1[|D z − D x| > b]`.
Telescoped, this turns the Tanaka occupation bound into
`window-occupation ≥ (E|D_T| − E|D_0| − ηT − 2R·∑_t farMass_t)/b`, so a per-step far mass summing to
`o(√T)` over the horizon leaves the `√T` Paley–Zygmund signal intact. Deterministic finite-sum.
-/
open Finset
open AnalyticCombinatorics AnalyticCombinatorics.Ch8 AnalyticCombinatorics.Ch8.Partitions AnalyticCombinatorics.Ch8.Partitions.Erdos
variable {ι : Type*} [Fintype ι] (K : ι → ι → ℝ) (D : ι → ℝ)
Formal statement
theorem AnalyticCombinatorics.Ch8.Partitions.Erdos.abs_drift_le_soft (b R : ℝ) (x : ι) (hbnn : 0 ≤ b)
(hKnn : ∀ z, 0 ≤ K x z) (hKsum : ∑ z, K x z = 1) (hR : ∀ z, |D z| ≤ R) :
(∑ z, K x z * |D z|) ≤ |D x| + b + 2 * R * farMass K D b x := by sorrySource
Human review
Confirmed by the mission captain (proposal self-audit).