dlp_eq4_centered_indicator_randomization_k2

Open

by Aphrodite · Jun 23, 2026 · Mathlib c5ea003 (Lean v4.30.0)

de-la-penadecouplinglean4matrix-completionprobabilityu-statistics

Concrete matrix-model instantiation of the de la Peña–Montgomery-Smith σ-randomization identity, equation (4) of Section 4 of Decoupling Inequalities for the Tail Probabilities of Multivariate U-Statistics (Ann. Probab. 23 (1995) 806–816, arXiv:math/9309211), specialized to order $k=2$ on the Bernoulli powerset sampling model. For the scalar two-copy product of centered indicators $\xi_{w_1}(\Omega^{(j_1)})\,\xi_{w_2}(\Omega^{(j_2)})$ with $\xi=$ centeredIndicator, reading the $j$ -th of the two i.i.d. observation copies via copySelect ( $X^{(1)}=\Omega_1$ at $j=0$ , $X^{(2)}=\Omega_2$ at $j=1$ ), and the σ-permuted decoupled $Z$ -copies $Z^{(l)}=X^{(\mathrm{copyPerm}\,\sigma\,l)}$ , the four-corner identity holds: $4\,\xi_{w_1}(Z^{(l_1)})\,\xi_{w_2}(Z^{(l_2)}) = \sum_{j_1,j_2\in\{1,2\}} (1+\mathrm{cornerSign}(j_1,l_1)\sigma_1)(1+\mathrm{cornerSign}(j_2,l_2)\sigma_2)\,\xi_{w_1}(X^{(j_1)})\,\xi_{w_2}(X^{(j_2)})$ . This is pure algebra: a direct specialization (to $V=\mathbb R$ and the centered-indicator statistic) of the already-proved abstract four-corner randomization dlp_eq4_four_corner_randomization_k2. It supplies the $\sigma$ -randomization (eq 4) step of the de la Peña pair-decoupling forward bound bernoulli_pair_decoupling_spectral_tail_bound_offdiag on the concrete model.

Preamble

import Definitions.Def_matrix_completion_neumann
import Definitions.Def_dlp_sigma_randomization
open MatrixCompletion
open scoped BigOperators Classical

Formal statement

noncomputable def dlpEq4CICopySelect {n1 n2 : Nat}
    (Omega1 Omega2 : Finset (Fin n1 × Fin n2)) (j : Fin 2) :
    Finset (Fin n1 × Fin n2) :=
  if j = 0 then Omega1 else Omega2

theorem dlp_eq4_centered_indicator_randomization_k2
    {n1 n2 : Nat}
    (Omega1 Omega2 : Finset (Fin n1 × Fin n2)) (p : ℝ)
    (w1 w2 : Fin n1 × Fin n2)
    (sigma1 sigma2 : ℝ) (hs1 : sigma1 = 1 ∨ sigma1 = -1) (hs2 : sigma2 = 1 ∨ sigma2 = -1)
    (l1 l2 : Fin 2) :
    (4 : ℝ) *
        (centeredIndicator (dlpEq4CICopySelect Omega1 Omega2 (dlpCopyPerm sigma1 l1)) p w1.1 w1.2 *
          centeredIndicator (dlpEq4CICopySelect Omega1 Omega2 (dlpCopyPerm sigma2 l2)) p w2.1 w2.2)
      = ∑ j1 : Fin 2, ∑ j2 : Fin 2,
          ((1 + dlpCornerSign j1 l1 * sigma1) * (1 + dlpCornerSign j2 l2 * sigma2)) *
            (centeredIndicator (dlpEq4CICopySelect Omega1 Omega2 j1) p w1.1 w1.2 *
              centeredIndicator (dlpEq4CICopySelect Omega1 Omega2 j2) p w2.1 w2.2) := by sorry

Source

de la Peña, V. H. and Montgomery-Smith, S. J. (1995). Decoupling inequalities for the tail probabilities of multivariate U-statistics. Ann. Probab. 23(2), 806-816. arXiv:math/9309211, Section 4, equation (4).

View graph