Open Access
2023 Spectral asymptotics for contracted tensor ensembles
Benson Au, Jorge Garza-Vargas
Author Affiliations +
Electron. J. Probab. 28: 1-32 (2023). DOI: 10.1214/23-EJP1001

Abstract

Let Td,N:ΩRNd be a random real symmetric Wigner-type tensor. For unit vectors (uN(i,j))iI,j[d2]SN1, we study the contracted tensor ensemble

1NTd,NuN(i,1)uN(i,d2)iI.

For large N, we show that the joint spectral distribution of this ensemble is well-approximated by a semicircular family (si)iI whose covariance (Ki,i(N))i,iI is given by the rescaled overlaps of the corresponding symmetrized contractions

Ki,i(N)=1d(d1)uN(i,1)uN(i,d2),uN(i,1)uN(i,d2),

which is the true covariance of the ensemble up to a Od(N1) correction. We further characterize the extreme cases of the variance Ki,i(N)[1d!,1d(d1)]. Our analysis relies on a tensorial extension of the usual graphical calculus for moment method calculations in random matrix theory, allowing us to access the independence in our random tensor ensemble.

Funding Statement

JGV was supported in part by NSF grant CCF-2009011.

Acknowledgments

The authors would like to thank the anonymous referees for their detailed feedback. In particular, the authors are grateful to the anonymous referee who suggested a shorter proof of Proposition 4.2 but insisted on remaining anonymous.

Citation

Download Citation

Benson Au. Jorge Garza-Vargas. "Spectral asymptotics for contracted tensor ensembles." Electron. J. Probab. 28 1 - 32, 2023. https://doi.org/10.1214/23-EJP1001

Information

Received: 19 May 2023; Accepted: 1 August 2023; Published: 2023
First available in Project Euclid: 5 October 2023

MathSciNet: MR4650897
Digital Object Identifier: 10.1214/23-EJP1001

Subjects:
Primary: 46L54 , 60B20
Secondary: 15B52 , 46L53

Keywords: Free probability , Random matrix , random tensor

Vol.28 • 2023
Back to Top