conditional_khintchine_scale_moment_from_row_column_energy_moment_2pN
ProvedConversion of the Bernoulli moment of the conditional Khintchine variance scale into the displayed log-moment bound, in the wider moment window (with , ). Given positive constants , there is such that for all , dimensions , , , exponent with and , and matrix : if the Bernoulli expectation of the -th power of the maximal row/column sampled energy is bounded by , then the Bernoulli expectation of the conditional Khintchine scale -moment is bounded by . This is the variant of Lemma 6.2's energy power-mean step required by the master conversion.
Preamble
import Definitions.Def_matrix_completion_rademacher open MatrixCompletion
Formal statement
theorem conditional_khintchine_scale_moment_from_row_column_energy_moment_2pN
(Cenergy Ckh : ℝ) :
0 < Cenergy →
0 < Ckh →
∃ Crad : ℝ, 0 < Crad ∧
∀ (β : ℝ), 2 < β →
∀ (n₁ n₂ m q : ℕ) (X : Matrix (Fin n₁) (Fin n₂) ℝ),
0 < n₁ → 0 < n₂ → m ≤ n₁ * n₂ →
(m : ℝ) ≥ β * (↑(max n₁ n₂)) *
Real.log (↑(max n₁ n₂)) →
1 ≤ q →
(q : ℝ) ≥ β * Real.log (↑(max n₁ n₂)) →
(q : ℝ) ≤ 2 * (β * Real.log (↑(max n₁ n₂))) →
(q : ℝ) ≤ 2 * (((m : ℝ) / ((n₁ : ℝ) * (n₂ : ℝ))) *
(↑(max n₁ n₂))) →
bernoulliExpectation ((m : ℝ) / ((n₁ : ℝ) * (n₂ : ℝ)))
(fun Omega =>
(max (sampledRowEnergyMax Omega X)
(sampledColumnEnergyMax Omega X)) ^ q) ≤
(Cenergy * ((m : ℝ) / ((n₁ : ℝ) * (n₂ : ℝ))) *
(↑(max n₁ n₂)) * entrySupNorm X ^ 2) ^ q →
bernoulliExpectation ((m : ℝ) / ((n₁ : ℝ) * (n₂ : ℝ)))
(fun Omega =>
(Ckh * Real.sqrt (q : ℝ) *
(((m : ℝ) / ((n₁ : ℝ) * (n₂ : ℝ)))⁻¹) *
Real.sqrt
(max (sampledRowEnergyMax Omega X)
(sampledColumnEnergyMax Omega X))) ^ q) ≤
(Crad * Real.sqrt
(((q : ℝ) * (↑(max n₁ n₂))) /
((m : ℝ) / ((n₁ : ℝ) * (n₂ : ℝ)))) *
entrySupNorm X) ^ q := by
sorrySource
Candes-Recht 2009 Exact Matrix Completion, Section 6.1, Lemma 6.2 (energy power-mean step), wide moment window q<=2(pN).