OAI/Probability/SLE/Gauge/CoverNumerics.lean

GitHub source file

6 declarations, 0 contain sorry, 0 declared as axioms

The failure and residual-cover estimates used by the explicit gauge.

  1. lemma OAI.SLEExactGauge.dense_cover_power_cancellation line 14

    Source documentation

    The dimensions in the hit probability, covering cost, and grid count cancel.

    lemma dense_cover_power_cancellation (κ : ℝ) {r : ℝ} (hr : 0 < r) :
        r ^ (-2 : ℝ) * r ^ dimension κ * r ^ codimension κ = 1 := by
    …

    Mentions: OAI.SLEExactGauge.dimension, OAI.SLEExactGauge.codimension

  2. lemma OAI.SLEExactGauge.failure_product_le_exp line 22
  3. theorem OAI.SLEExactGauge.failure_window_le_stretchedExp line 31

    Source documentation

    A linear number of tests, each succeeding with probability at least n^(-1/2), has the stretched-exponential failure bound used in the paper.

    theorem failure_window_le_stretchedExp {n : ℝ} (hn : 1 ≤ n)
        (N : ℕ) (hN : n / 4 ≤ (N : ℝ)) :
        (1 - 1 / Real.sqrt n) ^ N ≤ Real.exp (-Real.sqrt n / 4) := by
    …
  4. theorem OAI.SLEExactGauge.logarithmic_cleanup_tendsto_zero line 50

    Source documentation

    A logarithmic correction cannot offset the stretched-exponential failure probability obtained from the finite sequence of adapted tests.

    theorem logarithmic_cleanup_tendsto_zero {q c : ℝ}
        (hq : q ≤ 1) (hc : 0 < c) :
        Tendsto (fun x : ℝ => (Real.log x) ^ q * Real.exp (-c * Real.sqrt x))
          atTop (𝓝 0) := by
    …
  5. theorem OAI.SLEExactGauge.explicit_cleanup_tendsto_zero line 74
  6. theorem OAI.SLEExactGauge.expected_missed_cell_cost line 84

    Source documentation

    Finite expected residual cost is the sum of each missed-cell probability times its deterministic covering cost.

    theorem expected_missed_cell_cost
        {Ω ι : Type*} [MeasurableSpace Ω] (P : Measure Ω)
        (I : Finset ι) (missed : ι → Set Ω)
        (hm : ∀ i ∈ I, MeasurableSet (missed i))
        (cost bound : ι → ℝ≥0∞)
        (hb : ∀ i ∈ I, P (missed i) ≤ bound i) :
        (∫⁻ ω, ∑ i ∈ I, (missed i).indicator (fun _ => cost i) ω ∂P) ≤
          ∑ i ∈ I, cost i * bound i := by
    …