Tanaka Telescope
DefinitionAnalyticCombinatorics_Ch8_Partitions_TanakaTelescopeanalytic-combinatoricsfinite-kernel-formslean4partitions-remainder
The source mathematical objects and operations are specified by the following definitions.
def distPow (K : ι → ι → ℝ) (μ0 : ι → ℝ) : ℕ → ι → ℝ
| 0 => μ0
| (t + 1) => fun z => ∑ x, distPow K μ0 t x * K x z
Definition code
import Mathlib
set_option autoImplicit true
set_option relaxedAutoImplicit false
set_option maxSynthPendingDepth 3
/-!
# R7 Fact B via Doeblin: Tanaka telescoping ⟹ window occupation lower bound
Telescoping the single-step `|D|`-drift (brick 85) over the evolving distribution `distPow K μ₀ t`
(the law of the walk after `t` steps) yields, for a mean-preserving bounded-increment kernel,
`E|D_T| − E|D_0| ≤ b · ∑_{t<T} (window-occupation at step t)`,
i.e. `window-occupation ≥ (E|D_T| − E|D_0|)/b`. Combined with the Paley–Zygmund bound `E|D_T| ≥ c√T`
(brick 84) and the per-step second/fourth moments of the Erdős rank-difference walk, this gives the
occupation lower bound `≥ 1/δ` that closes the Hardy–Ramanujan convergence. All deterministic finite-sum
algebra. Opus-authored.
-/
noncomputable section
open Finset
namespace AnalyticCombinatorics.Ch8.Partitions.Erdos
variable {ι : Type*} [Fintype ι]
/-- Law of the walk after `t` steps from initial distribution `μ₀` under kernel `K`. -/
def distPow (K : ι → ι → ℝ) (μ0 : ι → ℝ) : ℕ → ι → ℝ
| 0 => μ0
| (t + 1) => fun z => ∑ x, distPow K μ0 t x * K x z
end AnalyticCombinatorics.Ch8.Partitions.Erdos
Source