Abstract
We compute the high-dimensional limit of the free energy associated with a multi-layer generalized linear model. Under certain technical assumptions, we identify the limit in terms of a variational formula. The approach is to first show that the limit is a solution to a Hamilton–Jacobi equation whose initial condition is related to the limiting free energy of a model with one fewer layer. Then, we conclude with an iteration.
Similar content being viewed by others
References
Barbier, J., Krzakala, F., Macris, N., Miolane, L., Zdeborová, L.: Optimal errors and phase transitions in high-dimensional generalized linear models. Proc. Natl. Acad. Sci. 116(12), 5451–5460 (2019)
Barbier, J., Macris, N.: The adaptive interpolation method: a simple scheme to prove replica formulas in Bayesian inference. Probab. Theory Relat. Fields 174(3–4), 1133–1185 (2019)
Barbier, J., Macris, N.: The adaptive interpolation method for proving replica formulas. Applications to the Curie–Weiss and Wigner spike models. J. Phys. A Math. Theor. 52(29), 294002 (2019)
Barbier, J., Macris, N., Miolane, L.: The layered structure of tensor estimation and its mutual information. In: 55th Annual Allerton Conference on Communication, Control, and Computing, pp. 1056–1063. IEEE, (2017)
Bardi, M., Evans, L.C.: On Hopf’s formulas for solutions of Hamilton–Jacobi equations. Nonlinear Anal. Theory Methods Appl. 8(11), 1373–1381 (1984)
Barra, A., Dal Ferraro, G., Tantari, D.: Mean field spin glasses treated with PDE techniques. Eur. Phys. J. B 86(7), 1–10 (2013)
Barra, A., Di Biasio, A., Guerra, F.: Replica symmetry breaking in mean-field spin glasses through the Hamilton–Jacobi technique. J. Stat. Mech. Theory Exp. 2010(09), P09006 (2010)
Boucheron, S., Lugosi, G., Massart, P.: Concentration Inequalities: A Nonasymptotic Theory of Independence. Oxford University Press, Oxford (2013)
Chen, H.-B.: Hamilton–Jacobi equations for nonsymmetric matrix inference. arXiv preprint arXiv:2006.05328 (2020)
Chen, H.-B., Mourrat, J.-C., Xia, J.: Statistical inference of finite-rank tensors. arXiv preprint arXiv:2104.05360 (2021)
Chen, H.-B., Xia, J.: Fenchel–Moreau identities on self-dual cones. arXiv preprint arXiv:2011.06979 (2020)
Chen, H.-B., Xia, J.: Hamilton–Jacobi equations for inference of matrix tensor products. arXiv preprint arXiv:2009.01678 (2020)
Dia, M., Macris, N., Krzakala, F., Lesieur, T., Zdeborová, L., et al.: Mutual information for symmetric rank-one matrix estimation: a proof of the replica formula. Adv. Neural Inf. Process. Syst. 29, 424–432 (2016)
Evans, L.C.: Partial Differential Equations, vol. 19. American Mathematical Society, Providence (2010)
Gabrié, M., Manoel, A., Luneau, C., Barbier, J., Macris, N., Krzakala, F., Zdeborová, L.: Entropy and mutual information in models of deep neural networks. J. Stat. Mech. Theory Exp. 2019(12), 124014 (2019)
Genovese, G., Barra, A.: A mechanical approach to mean field spin models. J. Math. Phys. 50(5), 053303 (2009)
Guerra, F.: Sum rules for the free energy in the mean field spin glass model. Fields Inst. Commun. 30(11), 161–170 (2001)
Lions, P.-L., Rochet, J.-C.: Hopf formula and multitime Hamilton–Jacobi equations. Proc. Am. Math. Soc. 96(1), 79–84 (1986)
Luneau, C., Barbier, J., Macris, N.: Mutual information for low-rank even-order symmetric tensor estimation. Inf. Inference A J. IMA 10(4), 1167–1207. arXiv:1904.04565 (2020). https://doi.org/10.1093/imaiai/iaaa022
Luneau, C., Macris, N., Barbier, J.: High-dimensional rank-one nonsymmetric matrix decomposition: the spherical case. arXiv preprint arXiv:2004.06975 (2020)
Mourrat, J.-C.: Free energy upper bound for mean-field vector spin glasses. arXiv preprint arXiv:2010.09114 (2020)
Mourrat, J.-C.: Hamilton–Jacobi equations for mean-field disordered systems. Ann. Henri Lebesgue 4, 453–484 (2021)
Mourrat, J.-C.: Nonconvex interactions in mean-field spin glasses. Probab. Math. Phys. 2(2), 61–119 (2021)
Mourrat, J.-C.: The Parisi formula is a Hamilton–Jacobi equation in Wasserstein space. Can. J. Math. 3, 1–23 (2021)
Mourrat, J.-C.: Hamilton–Jacobi equations for finite-rank matrix inference. Ann. Appl. Probab. 30(5), 2234–2260 (2020). https://doi.org/10.1214/19-AAP1556. https://projecteuclid.org/journals/annals-of-applied-probability/volume-30/issue-5/HamiltonJacobi-equations-for-finite-rank-matrix-inference/10.1214/19-AAP1556.short@article%7B10.1214/19-AAP1556
Mourrat, J.-C., Panchenko, D.: Extending the Parisi formula along a Hamilton–Jacobi equation. Electron. J. Probab. 25, 1–17 (2020)
Reeves, G.: Information-theoretic limits for the matrix tensor product. IEEE J. Sel. Areas Inf. Theory 1(3), 777–798 (2020)
Rockafellar, R.T.: Convex Analysis, vol. 36. Princeton University Press, Princeton (1970)
Acknowledgements
We warmly thank Jean-Christophe Mourrat for many helpful comments and discussions. We thank the referees for the comments that help us improve the paper considerably.
Author information
Authors and Affiliations
Corresponding author
Additional information
Communicated by S. Chatterjee.
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
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
Chen, HB., Xia, J. Free Energy of Multi-Layer Generalized Linear Models. Commun. Math. Phys. 400, 1861–1913 (2023). https://doi.org/10.1007/s00220-022-04630-4
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00220-022-04630-4