Nous considérons un modèle d’inférence statistique général de produits tensoriels de rang fini. Pour toute structure d’interaction et tout ordre de produits tensoriels, nous identifions l’énergie libre limite du modèle en termes d’une formule variationnelle. Notre approche consiste à montrer d’abord que l’énergie libre limite doit être la solution de viscosité d’une certaine équation de Hamilton–Jacobi.
We consider a general statistical inference model of finite-rank tensor products. For any interaction structure and any order of tensor products, we identify the limit free energy of the model in terms of a variational formula. Our approach consists of showing first that the limit free energy must be the viscosity solution to a certain Hamilton–Jacobi equation.
Révisé le :
Accepté le :
Publié le :
Mots clés : inference problem, Hamilton–Jacobi equation, tensor
@article{AHL_2022__5__1161_0, author = {Chen, Hongbin and Mourrat, Jean-Christophe and Xia, Jiaming}, title = {Statistical inference of finite-rank tensors}, journal = {Annales Henri Lebesgue}, pages = {1161--1189}, publisher = {\'ENS Rennes}, volume = {5}, year = {2022}, doi = {10.5802/ahl.146}, language = {en}, url = {http://www.numdam.org/articles/10.5802/ahl.146/} }
TY - JOUR AU - Chen, Hongbin AU - Mourrat, Jean-Christophe AU - Xia, Jiaming TI - Statistical inference of finite-rank tensors JO - Annales Henri Lebesgue PY - 2022 SP - 1161 EP - 1189 VL - 5 PB - ÉNS Rennes UR - http://www.numdam.org/articles/10.5802/ahl.146/ DO - 10.5802/ahl.146 LA - en ID - AHL_2022__5__1161_0 ER -
Chen, Hongbin; Mourrat, Jean-Christophe; Xia, Jiaming. Statistical inference of finite-rank tensors. Annales Henri Lebesgue, Tome 5 (2022), pp. 1161-1189. doi : 10.5802/ahl.146. http://www.numdam.org/articles/10.5802/ahl.146/
[ACCM21] The solution of the deep Boltzmann machine on the Nishimori line, Commun. Math. Phys., Volume 387 (2021) no. 2, pp. 1191-1214 | DOI | MR | Zbl
[BDBG10] Replica symmetry breaking in mean-field spin glasses through the Hamilton–Jacobi technique, J. Stat. Mech. Theory Exp., Volume 2010 (2010) no. 09, P09006 | MR | Zbl
[BDFT13] Mean field spin glasses treated with PDE techniques, Eur. Phys. J. B, Volume 86 (2013) no. 7, 332 | DOI | MR
[BDM + 16] Mutual information for symmetric rank-one matrix estimation: A proof of the replica formula, Advances in Neural Information Processing Systems (NIPS) (Lee, D.; Sugiyama, M.; Luxburg, U.; Guyon, I.; Garnett, R., eds.) (NeurIPS Proceedings), Volume 29, Curran Associates, Inc. (2016), pp. 424-432
[BM19a] The adaptive interpolation method: a simple scheme to prove replica formulas in Bayesian inference, Probab. Theory Relat. Fields, Volume 174 (2019) no. 3-4, pp. 1133-1185 | DOI | MR | Zbl
[BM19b] The adaptive interpolation method for proving replica formulas. Applications to the Curie–Weiss and Wigner spike models, J. Phys. A, Math. Theor., Volume 52 (2019) no. 29, 294002 | DOI | MR
[BMM17] The layered structure of tensor estimation and its mutual information, 2017 Annual Allerton Conference on Communication, Control, and Computing (Allerton), IEEE (2017), pp. 1056-1063 | DOI
[Che20] Hamilton–Jacobi equations for nonsymmetric matrix inference (2020) (https://arxiv.org/abs/2006.05328)
[CX20] Hamilton–Jacobi equations for inference of matrix tensor products (2020) (https://arxiv.org/abs/2009.01678)
[CX22] Hamilton–Jacobi equations from mean-field spin glasses (2022) (https://arxiv.org/abs/2201.12732)
[GB09] A mechanical approach to mean field spin models, J. Math. Phys., Volume 50 (2009) no. 5, 053303 | MR | Zbl
[Gue01] Sum rules for the free energy in the mean field spin glass model, Fields Institute Communications, Volume 30 (2001) no. 11 | MR | Zbl
[KG18] Statistical mechanics of low-rank tensor decomposition, Advances in Neural Information Processing Systems (Bengio, S.; Wallach, H.; Larochelle, H.; Grauman, K.; Cesa-Bianchi, N.; Garnett, R., eds.) (NeurIPS Proceedings), Curran Associates, Inc. (2018), pp. 8201-8212
[LBM21] Mutual information for low-rank even-order symmetric tensor estimation, Inf. Inference, Volume 10 (2021) no. 4, pp. 1167-1207 | DOI | MR | Zbl
[LM19] Fundamental limits of symmetric low-rank matrix estimation, Probab. Theory Relat. Fields, Volume 173 (2019) no. 3-4, pp. 859-929 | DOI | MR | Zbl
[LMB20] High-dimensional rank-one nonsymmetric matrix decomposition: the spherical case (2020) (https://arxiv.org/abs/2004.06975)
[LML + 17] Statistical and computational phase transitions in spiked tensor estimation, 2017 IEEE International Symposium on Information Theory (ISIT), IEEE (2017), pp. 511-515 | DOI
[Mio17] Fundamental limits of low-rank matrix estimation: the non-symmetric case (2017) (https://arxiv.org/abs/1702.00473)
[Mou20] Hamilton–Jacobi equations for finite-rank matrix inference, Ann. Appl. Probab., Volume 30 (2020) no. 5, pp. 2234-2260 | DOI | MR | Zbl
[Mou21a] Free energy upper bound for mean-field vector spin glasses (2021) (https://arxiv.org/abs/2010.09114)
[Mou21b] Hamilton–Jacobi equations for mean-field disordered systems, Ann. Henri Lebesgue, Volume 4 (2021), pp. 453-484 | DOI | MR | Zbl
[Mou21c] Nonconvex interactions in mean-field spin glasses, Probability and Mathematical Physics, Volume 2 (2021) no. 2, pp. 281-339 | DOI | MR | Zbl
[Mou22] The Parisi formula is a Hamilton–Jacobi equation in Wasserstein space, Can. J. Math., Volume 74 (2022) no. 3, pp. 607-629 | DOI | MR | Zbl
[MP20] Extending the Parisi formula along a Hamilton–Jacobi equation, Electron. J. Probab., Volume 25 (2020), 23 | DOI | MR | Zbl
[MR19] Mutual information in community detection with covariate information and correlated networks, 2019 Annual Allerton Conference on Communication, Control, and Computing (Allerton), IEEE (2019), pp. 602-607 | DOI
[Ree20] Information-Theoretic Limits for the Matrix Tensor Product (2020) (https://arxiv.org/abs/2005.11273)
[RMV19] The geometry of community detection via the mmse matrix, 2019 IEEE International Symposium on Information Theory (ISIT), IEEE (2019), pp. 400-404 | DOI
[Roc70] Convex Analysis, Princeton Mathematical Series, 36, Princeton University Press, 1970 | DOI | Zbl
[Zha11] Matrix theory, Universitext, Springer, 2011 | DOI | MR | Zbl
Cité par Sources :