Universality of the Wigner-Gurau limit for random tensors
Abstract: In this article, we develop a combinatorial approach for studying moments of the resolvent trace for random tensors proposed by Razvan Gurau. Our work is based on the study of hypergraphs and extends the combinatorial proof of moments convergence for Wigner's theorem. This also opens up paths for research akin to free probability for random tensors. Specifically, trace invariants form a complete basis of tensor invariants and constitute the moments of the resolvent trace. For a random tensor with entries independent, centered, with the right variance and bounded moments, we prove the convergence of the expectation and bound the variance of the balanced single trace invariant. This implies the universality of the convergence of the associated measure towards the law obtained by Gurau in the Gaussian case, whose limiting moments are given by the Fuss-Catalan numbers. This generalizes Wigner's theorem for random tensors. Additionally, in the Gaussian case, we show that the limiting distribution of the $k$-times contracted $p$-order random tensor by a deterministic vector is always the Wigner-Gurau law at order $p-k$, dilated by $\sqrt{\binom{p-1}{k}}$. This establishes a connection with the approach of random tensors through the matrix study of the $p-2$ contractions of the tensor.
- L. Qi “Eigenvalues of a real supersymmetric tensor” In Journal of Symbolic Computation 40, 2005
- T.Banica S.Belinschi M.Capitaine B. Collins “Free Bessel laws”, 2008 URL: https://arxiv.org/pdf/0710.5931.pdf
- D.Cartwright B. Sturmfels “The number of eigenvalues of a tensor”, 2010 URL: https://arxiv.org/pdf/1004.4953.pdf
- K.A.Penson K. Zyczkowski “Product of ginibre matrices: Fuss-catalan and raney distributions” In Physical Review 83, 2011
- J.M.Landsberg Y.Qi K. Ye “On the geometry of tensor network states”, 2012 URL: https://arxiv.org/pdf/1105.4449.pdf
- A.Anandkumar R.Ge D.Hsu S.Kakade M. Telgarsky “Tensor decompositions for learning latent variable models”, 2014 URL: https://arxiv.org/pdf/1210.7559.pdf
- T.Kawano D.Obster N. Sasakura “Canonical tensor model through data analysis - Dimensions, topologies, and geometries -”, 2015 URL: https://arxiv.org/pdf/1506.04872.pdf
- B. Simon “A Comprehensive Course in Analysis” American Mathematical Society, 2015
- R. Gurau “Random tensors” Oxford University Press, 2016
- N.D.Sidiropoulos L.De Lathauwer X.Fu K.Huang E.E.Papalexakis C. Faloutsos “Tensor Decomposition for Signal Processing and Machine Learning” In IEEE Transactions on Signal Processing 65, 2017
- R. Gurau “Quenched equals annealed at leading order in the colored SYK model”, 2017 URL: https://arxiv.org/pdf/1702.04228.pdf
- N.Sasakura Y. Sato “Constraint algebra of general relativity from a formal continuum limit of canonical tensor model”, 2018 URL: https://arxiv.org/pdf/1805.04800.pdf
- P. Breiding “How many eigenvalues of a random symmetric tensor are real?”, 2019 URL: https://arxiv.org/pdf/1701.07312.pdf
- A.Jagannath P.Lopatto L. Miolane “Statistical thresholds for tensor PCA”, 2019 URL: https://arxiv.org/pdf/1812.03403.pdf
- R. Gurau “On the generalization of the Wigner semicircle law to real symmetric tensors model”, 2020 URL: https://arxiv.org/pdf/2004.02660.pdf
- J.-H.Goulart P.Comon R. Couillet “A random matrix perspective on random tensors”, 2022 URL: https://arxiv.org/pdf/2108.00774.pdf
- B.Collins R.Gurau L. Lionni “The tensor Harish-Chandra–Itzykson–Zuber integral II: detecting entanglement in large quantum systems”, 2022 URL: https://arxiv.org/pdf/2201.12778.pdf
- B.Collins R.Gurau L. Lionni “The tensor Harish-Chandra-Itzykson-Zuber integral I: Weingarten calculus and a generalization of monotone Hurwitz numbers”, 2022 URL: https://arxiv.org/pdf/2010.13661.pdf
- M.Seddik M.Tiomoko A.Decurninge M.Panov M. Guillaud “Learning from Low Rank Tensor Data: A Random Tensor Theory Perspective” In Proceedings of Machine Learning Research 216, 2023
Paper Prompts
Sign up for free to create and run prompts on this paper using GPT-5.
Top Community Prompts
Collections
Sign up for free to add this paper to one or more collections.
