Abstract
We obtain estimates for the Kolmogorov distance to appropriately chosen gaussians, of linear functions
of random tensors \(\varvec{X}=\langle X_i:i\in [n]^d\rangle \) which are symmetric and exchangeable, and whose entries have bounded third moment and vanish on diagonal indices. These estimates are expressed in terms of intrinsic (and easily computable) parameters associated with the random tensor \(\varvec{X}\) and the given coefficients \(\langle \theta _i:i\in [n]^d\rangle \), and they are optimal in various regimes. The key ingredient—which is of independent interest—is a combinatorial CLT for high-dimensional tensors which provides quantitative non-asymptotic normality under suitable conditions, of statistics of the form
where \(\varvec{\zeta }:[n]^d\times [n]^d\rightarrow \mathbb {R}\) is a deterministic real tensor, and \(\pi \) is a random permutation uniformly distributed on the symmetric group \(\mathbb {S}_n\). Our results extend, in any dimension d, classical work of Bolthausen who covered the one-dimensional case, and more recent work of Barbour/Chen who treated the two-dimensional case.
Similar content being viewed by others
Notes
In this context, the euclidean norm is also referred to as the Hilbert–Schmidt norm.
See Paragraph 12.2.1 for the definition of extendability.
Note that \(\pi \circ i:=\big (\pi (i_1),\dots ,\pi (i_s)\big )\in [n]^s_{\textrm{Inj}}\) for every \(\pi \in \mathbb {S}_n\) and every \(i=(i_1,\dots ,i_s)\in [n]^s_{\textrm{Inj}}\).
This result was not explicitly isolated in [3], but it follows fairly straightforwardly from the methods developed therein.
Recall that a random vector \(\varvec{X}\) in \(\mathbb {R}^n\) is called isotropic if its entries are uncorrelated random variables with zero mean and unit variance.
We note that closely related questions have been studied earlier—see, e.g. [46].
Here, we identify \(\left( {\begin{array}{c}[n]\\ k\end{array}}\right) \) with the set of all \(x\in \{0,1\}^n\) which have exactly k nonzero coordinates.
We follow the convention that \(\frac{1}{0}=\infty \).
That is, t is either the empty map, or a nonempty, partial, one-to-one map from [s] into [n].
That is, \(t(i_1,i_2)\) is the unique permutation which maps \(i_1\) to \(i_2\) and \(i_2\) to \(i_1\), and it is the identity on \([n]\setminus \{i_1,i_2\}\).
More precisely, in these cases the index set I can be identified with the set \(\left( {\begin{array}{c}[n]\\ d\end{array}}\right) \) and the group G consists of those permutations \(\overline{\pi }\) of \(\left( {\begin{array}{c}[n]\\ d\end{array}}\right) \) which are of the form \(\overline{\pi }\big (\{j_1,\dots ,j_d\}\big )= \{\pi (j_1),\dots ,\pi (j_d)\}\) for some permutation \(\pi \) of [n].
In fact, it is quite likely that this question has already been asked, but we could not find something relevant in the literature.
That is, \(\sum _{i_0\sqsubseteq i\in [n]^s} g_s(i)=0\) for every \(i_0\in [n]^{s-1}\).
Here, for every tensor \(g:[n]^s\rightarrow \mathbb {R}\) and every \(p>1\) we set \(\Vert g\Vert _{\ell _p}:=\big ( \sum _{i\in [n]^s} |g(i)|^p\big )^{1/p}\).
Specifically, it affects the constant \(\kappa \).
Recall that for every \(i=(i_1,\dots ,i_r)\in [n]^r\) and every \(\tau \in \mathbb {S}_r\) we set \(i \circ \tau =(i_{\tau (1)},\dots ,i_{\tau (r)})\in [n]^r\).
Here, given two positive quantities a and b, we write \(a\gtrsim b\) to denote the fact that \(a\geqslant C b\) for some positive universal constant C.
More precisely, of length cn where \(c>0\) is a sufficiently small constant.
Here, we assume that this process is sufficiently measurable.
The argument in this context is actually simpler.
References
Aldous, D.J.: Representations for partially exchangeable arrays of random variables. J. Multivar. Anal. 11, 581–597 (1981)
Aldous, D.J.: Exchangeability and related topics. In: École d’été de Probabilités de Saint-Flour XIII, Lecture Notes in Mathematics, vol. 1117, pp. 1–198. Springer (1983)
Barbour, A.D., Chen, L.H.Y.: The permutation distribution of matrix correlation statistics. In: Stein’s Method and Applications, Lecture Notes Series, vol. 5, pp. 223–245. Institute for Mathematical Sciences, National University of Singapore, Singapore University Press (2005)
Barbour, A.D., Chen, L.H.Y.: Stein’s (Magic) Method (2014). https://arxiv.org/pdf/1411.1179.pdf
Barbour, A.D., Eagleson, G.K.: Random association of symmetric arrays. Stoch. Anal. Appl. 4, 239–281 (1986)
Blum, J.R., Chernoff, H., Rosenblatt, H., Teicher, H.: Central limit theorems for interchangeable processes. Canada J. Math. 10, 222–229 (1958)
Bobkov, S.G.: Concentration of normalized sums and a central limit theorem for noncorrelated random variables. Ann. Probab. 32, 2884–2908 (2004)
Bobkov, S.G., Chistyakov, G.P., Götze, F.: Berry–Esseen bounds for typical weighted sums. Electron. J. Probab. 23, 1–22 (2018)
Bolthausen, E.: An estimate of the remainder in a combinatorial central limit theorem. Z. Wahrsch. Verw. Gebiete 66, 379–386 (1984)
Bolthausen, E., Götze, F.: The rate of convergence for multivariate sampling statistics. Ann. Stat. 21, 1692–1710 (1993)
Bloznelis, M., Götze, F.: Orthogonal decomposition of finite population statistics and its distributional asymptotics. Ann. Stat. 29, 899–917 (2001)
Bloznelis, M., Götze, F.: An Edgeworth expansion for symmetric finite population statistics. Ann. Probab. 30, 1238–1265 (2002)
Carbery, A., Wright, J.: Distributional and \(L^q\) norm inequalities for polynomials over convex bodies in \(\mathbb{R} ^n\). Math. Res. Lett. 8, 233–248 (2001)
Chatterjee, S.: A short survey of Stein’s method. In: Proceedings of the International Congress of Mathematicians, vol. IV, pp. 13–21, Seoul, Korea (2014)
Chen, L.H.Y., Fang, X.: On the error bound in a combinatorial central limit theorem. Bernoulli 21, 335–359 (2015)
Chen, L.H.Y., Goldstein, L., Shao, Q.M.: Normal Approximation by Stein’s Method. Probability and Its Applications. Springer, Berlin (2011)
Chen, L.H.Y., Shao, Q.M.: Normal approximation for nonlinear statistics using a concentration inequality approach. Bernoulli 13, 581–599 (2007)
Costello, K.P., Tao, T., Vu, V.: Random symmetric matrices are almost surely nonsingular. Duke Math. J. 135, 395–413 (2006)
Daniels, H.E.: The relation between measures of correlation in the universe of sample permutations. Biometrika 33, 129–135 (1944)
Davezies, L., D’Haultfœuille, X., Guyonvarch, Y.: Empirical process results for exchangeable arrays. Ann. Stat. 49, 845–862 (2021)
de Jong, P.: A central limit theorem for generalized multilinear forms. J. Multivar. Anal. 34, 275–289 (1990)
Diaconis, P., Freedman, D.: Finite exchangeable sequences. Ann. Probab. 8, 745–764 (1980)
Diaconis, P., Freedman, D.: A dozen de Finetti-style results in search of a theory. Ann. Inst. Henri Poincaré Probab. Stat. 23, 397–423 (1987)
Döbler, Ch., Kasprzak, M., Peccati, G.: The multivariate functional de Jong CLT. Probab. Theory Relat. Fields 184, 367–399 (2022)
Eagleson, G.K., Weber, N.C.: Limit theorems for weakly exchangeable arrays. Math. Proc. Camb. Philos. Soc. 84, 123–130 (1978)
Erdős, P.: On a lemma of Littlewood and Offord. Bull. Am. Math. Soc. 51, 898–902 (1945)
Filmus, Y., Kindler, G., Mossel, E., Wimmer, K.: Invariance principle on the slice. ACM Trans. Comput. Theory 10, 37 (2018)
Filmus, Y., Mossel, E.: Harmonicity and invariance on slices of the Boolean cube. Probab. Theory Relat. Fields 175, 721–782 (2019)
Graham, R.L., Knuth, D.E., Patashnik, O.: Concrete Mathematics: A Foundation for Computer Science, 2nd edn. Addison-Wesley Publishing Group, Boston (1994)
Hoeffding, W.: A class of statistics with asymptotically normal distribution. Ann. Math. Stat. 19, 293–325 (1948)
Hoeffding, W.: A combinatorial central limit theorem. Ann. Math. Stat. 22, 558–566 (1951)
Hoover, D.N.: Relations on probability spaces and arrays of random variables (1979). https://www.stat.berkeley.edu/~aldous/Research/hoover.pdf
Kallenberg, O.: Probabilistic Symmetries and Invariance Principles. Probability and its Applications. Springer, New York (2005)
Kwan, M., Sudakov, B., Tran, T.: Anticoncentration for subgraph statistics. J. Lond. Math. Soc. 99, 757–777 (2019)
Littlewood, J.E., Offord, A.C.: On the number of real roots of a random algebraic equation. III. Rec. Math. (Mat. Sbornik) Nouv. Ser. 12, 277–286 (1943)
Litvak, A.E., Lytova, A., Tikhomirov, K., Tomczak-Jaegermann, N., Youssef, P.: Adjacency matrices of random digraphs: singularity and anti-concentration. J. Math. Anal. Appl. 445, 1447–1491 (2017)
Loh, W.L.: On Latin hypercube sampling. Ann. Stat. 24, 2058–2080 (1996)
Meka, R., Nguyen, O., Vu, V.: Anti-concentration for polynomials of independent random variables. Theory Comput. 12, Paper No. 11, 16 (2016)
Mossel, E., O’Donnell, R., Oleszkiewicz, K.: Noise stability of functions with low influences: invariance and optimality. Ann. Math. 171, 295–341 (2010)
Nguyen, H.H., Vu, V.: Small ball probability, inverse theorems, and applications. In: Erdős centennial, Bolyai Society Mathematical Studies, vol. 25, pp. 409–463. Springer, Berlin (2013)
Nourdin, I., Peccati, G.: Normal Approximations with Malliavin Calculus. From Stein’s Method to Universality, Cambridge Tracts in Mathematics, vol. 192. Cambridge University Press, Cambridge (2012)
Peccati, G.: Hoeffding-ANOVA decompositions for symmetric statistics of exchangeable observations. Ann. Probab. 32, 1796–1829 (2004)
Peccati, G., Pycke, J.-R.: Hoeffding spaces and Specht modules. ESAIM Probab. Stat. 15, S58–S68 (2011)
Razborov, A., Viola, E.: Real advantage. ACM Trans. Comput. Theory 5, 1–8 (2013)
Robbins, H.: A remark on Stirling’s formula. Am. Math. Mon. 62, 26–29 (1955)
Rosiński, J., Samorodnitsky, G.: Symmetrization and concentration inequalities for multilinear forms with applications to zero-one laws for Lévy chaos. Ann. Probab. 24, 422–437 (1996)
Serfling, R.J.: Approximation Theorems of Mathematical Statistics. Wiley Series in Probability and Mathematical Statistics. Wiley, Hoboken (1980)
Silverman, B.W.: Limit theorems for dissociated random variables. Adv. Appl. Probab. 8, 806–819 (1976)
Stein, C.: A bound for the error in the normal approximation to the distribution of a sum of dependent random variables. In: Proceedings of the Sixth Berkeley Symposium on Mathematical Statistics and Probability, vol. II: Probability Theory, pp. 583–602. Univ. California Press (1982)
Stein, C.: Approximate Computation of Expectations, Institute of Mathematical Statistics Lecture Notes—Monograph Series, vol. 7. Hayward, CA (1986)
Vu, V.: Concentration and Anti-concentration. https://www.ima.umn.edu/2014-2015/W9.8-12.14/21252
Wald, A., Wolfowitz, J.: Statistical tests based on permutation of the observations. Ann. Math. Stat. 15, 358–372 (1944)
Zhao, L., Chen, X.: Normal approximation for finite-population U-statistics. Acta Math. Appl. Sin. 6, 263–272 (1990)
Zhao, L., Bai, Z., Chao, C.-C., Liang, W.-Q.: Error bound in a central limit theorem of double-indexed permutation statistics. Ann. Stat. 25, 2210–2227 (1997)
Acknowledgements
We would like to thank the anonymous referees for their comments, remarks and suggestions. The research was supported by the Hellenic Foundation for Research and Innovation (H.F.R.I.) under the “2nd Call for H.F.R.I. Research Projects to support Faculty Members & Researchers” (Project Number: HFRI-FM20-02717).
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix A. Proof of Lemma 7.3
Appendix A. Proof of Lemma 7.3
The argument is an interesting and nontrivial instance of the Stein/Chen method of normal approximation via concentration (see, e.g., [4, Section 6]).
A.1. Recall that n, d are positive integers with \(d\geqslant 2\) and \(n\geqslant 4d^2\). We start by setting
Notice that, by (7.6), we have
Using (7.5), (7.6), (A.2) and the fact that \(\min (\alpha ,\beta )\geqslant \alpha - \frac{\alpha ^2}{4\beta }\) for every \(\alpha \geqslant 0\) and \(\beta >0\), we obtain that
Sublemma A.1 1
We have
We postpone the proof of Sublemma A.1 to the end of this appendix.
A.2. Now let \(z\in \mathbb {R}\) be arbitrary. We will show that the probabilities \({\mathbb {P}}(z-|\Theta |\leqslant \Xi _1\leqslant z)\) and \({\mathbb {P}}\big (z\leqslant \Xi _1 \leqslant z+|\Theta |\big )\) are both upper-bounded by the quantity appearing in the right-hand-side of (7.20); clearly, this is enough to complete the proof. The argument is symmetric, and so we shall focus on bounding the probability \({\mathbb {P}}\big (z\leqslant \Xi _1 \leqslant z+|\Theta |\big )\).
Let \(f_\Theta :\mathbb {R}\rightarrow \mathbb {R}\) be defined by
By (\(\mathcal {E}\)4), the pair \((\pi _1,\pi _2)\) is exchangeable. Hence, by (7.1) and (7.19), we have
which in turn implies, after adding \(2\,\mathbb {E}\big [(\Xi _1-\Xi _1')f_\Theta (\Xi _1)\big ]\) and multiplying by \(\frac{n}{4}\), that
Moreover, by (7.4), we have
and so, by (7.3), the Cauchy–Schwarz inequality and the fact that \(\Vert f_\Theta \Vert _{L_\infty }\leqslant \delta +\frac{1}{2}|\Theta |\),
By the definition of \(\Theta \) in (7.19), the triangle inequality and (7.7), we thus have
Next observe that \(\Vert f_\Theta - f_{\Theta '}\Vert _{L_{\infty }}\leqslant \frac{1}{2}\,\big ||\Theta |-|\Theta '|\big | \leqslant \frac{1}{2}\,|\Theta -\Theta '|\). Hence,
where we have used the Cauchy–Schwarz inequality, the triangle inequality and (7.8).
From this point on, the proof proceeds exactly as in [3, Lemma 2.2]. More precisely, in order to estimate the second term of the right-hand-side of (A.7), by the fundamental theorem of calculus, we have
with the convention that \(\textbf{1}_{[\beta ,\alpha ]}\) is constantly 0 if \(\alpha <\beta \). Also observe that, by (A.5), we have \(f'_\Theta =\textbf{1}_{[z-\delta ,z+\delta +|\Theta |]}\) which implies that for every \(t\in \mathbb {R}\),
-
\((\Xi _1'-\Xi _1)\,f'_\Theta (\Xi _1+t)\, \big (\textbf{1}_{[0,\Xi _1'-\Xi _1]}(t)-\textbf{1}_{[\Xi _1'-\Xi _1,0]}(t)\big ) \geqslant 0\).
Therefore,
Moreover, for every \(|t|\leqslant \delta \) we have
-
\((\Xi _1'-\Xi _1)\, \big (\textbf{1}_{[0,\Xi _1'-\Xi _1]}(t)-\textbf{1}_{[\Xi _1'-\Xi _1,0]}(t)\big )\geqslant 0\) and
-
\(\textbf{1}_{[z-\delta ,z+\delta +|\Theta |]}(\Xi _1+t) \geqslant \textbf{1}_{[z,z+|\Theta |]}(\Xi _1)\),
and so, by (A.13),
On the other hand, by Sublemma 1 and inequality \(\alpha \beta \leqslant \frac{1}{2}\,(\alpha ^2+\beta ^2)\) applied for “\(\alpha =\frac{1}{\sqrt{2}}\,\textbf{1}_{[z,z+|\Theta |]}(\Xi _1)\)" and “\(\beta = \sqrt{2}\,(\eta _\delta -\mathbb {E}[\eta _\delta ])\)", we see that
Thus, by (A.14) and (A.15), we have
Invoking the choice of \(\delta \) in (A.1) and the choice of A in (A.12) and combining (A.7), (A.10), (A.11) and (A.16), we conclude that
1.1 A.3 Proof of Sublemma A.1
Let \(\varvec{\zeta }:[n]^2\times [n]^2\rightarrow \mathbb {R}\) be defined by setting
for every \(i,j,p,q\in [n]\) with \(i\ne j\) and \(p\ne q\); otherwise, set \(\varvec{\zeta }(i,j,p,q)=0\). Then observe that, by the choice of \(\eta _{\delta }\) in (A.1), we have
Also recall that, by (\(\mathcal {E}\)3), \(\pi _1\) is uniformly distributed on \(\mathbb {S}_n\). Thus, (A.19) asserts the random variable \(\eta _\delta \) can be expressed as the Z-statistic of order two associated with the tensor \(\varvec{\zeta }\) which has a rather special form.
The proof is based on a specific decomposition of this Z-statistic. This decomposition, also introduced by Barbour and Chen [3], is less symmetric than the decomposition described in Sect. 9 (yet it has enough symmetries in order to be computationally useful), but it has the advantage that its non-constant components are mean zero random variables. Of course, this property is very useful for computing the variance. For the convenience of the reader we shall briefly recall this decomposition; for more information we refer to [3].
Specifically, for every \(i,j,p,q\in [n]\) set
Then, for every \(i,j,p,q\in [n]\) we define
Notice that for every \(i,p,q\in [n]\) we have \(\varvec{\zeta }_D(i,i,p,q)=0\); moreover, for every \(i,j\in [n]\),
By (A.19) and the previous definitions, we may decompose \(\eta _\delta \) as
Invoking (A.23) and the fact that \(\pi _1\) is uniformly distributed on \(\mathbb {S}_n\) we see, in particular, that \(\mathbb {E}[\eta _\delta ] = \frac{n-1}{4}\varvec{\zeta }(\cdot ,\cdot ,\cdot ,\cdot )\). Therefore,
Finally, using the Cauchy–Schwarz inequality, (A.23) and arguingFootnote 20 as in Sect. 5.6, it is not hard to verify that
By (A.25), (A.26) and the fact that \(n\geqslant 3\), we conclude that (A.4) is satisfied.
Rights and permissions
Springer Nature or its licensor (e.g. a society or other partner) holds exclusive rights to this article under a publishing agreement with the author(s) or other rightsholder(s); author self-archiving of the accepted manuscript version of this article is solely governed by the terms of such publishing agreement and applicable law.
About this article
Cite this article
Dodos, P., Tyros, K. Anticoncentration and Berry–Esseen bounds for random tensors. Probab. Theory Relat. Fields 187, 317–384 (2023). https://doi.org/10.1007/s00440-023-01211-x
Received:
Revised:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00440-023-01211-x
Keywords
- Random tensors
- Exchangeability
- Berry–Esseen bounds
- Anticoncentration
- Combinatorial central limit theorem
- Stein’s method
- Polynomials of boolean random variables