tropp_even_schatten_variance_scale
Provedconcentrationkhintchinematrix-completionschatten
Tropp-route even-Schatten variance-scale moment bound (noncommutative Khintchine via the Tropp/Buchholz matrix moment recursion, bypassing Lust-Picquard duality). For a Rademacher-symmetrized sampled matrix and even exponent (), the Rademacher expectation of is at most , where is the Buchholz pairing-count (double-factorial) constant. This is the trace-moment self-bounding recursion giving the Khintchine factor at even integer order.
Preamble
import Definitions.Def_matrix_completion_schatten import Definitions.Def_matrix_completion_rademacher open MatrixCompletion
Formal statement
theorem tropp_even_schatten_variance_scale {n1 n2 : ℕ}
(n : ℕ) (hn : 1 ≤ n)
(Omega : Finset (Fin n1 × Fin n2)) (p : ℝ) (hp : 0 < p)
(X : RealMatrix n1 n2) :
rademacherExpectation
(fun eps => schattenNorm (2 * n : ℝ) (rademacherSampledMatrix Omega eps p X) ^ (2 * n))
≤ ((Nat.factorial (2 * n) : ℝ) / ((2 ^ n : ℝ) * (Nat.factorial n : ℝ)))
* (rademacherSampledVarianceScale Omega p X) ^ (2 * n)
* ((n1 : ℝ) + (n2 : ℝ)) := by
sorrySource
Tropp, User-friendly tail bounds for matrix sums; CR2008 Section 6.1; Buchholz Math. Ann. 2001