Abstract
We prove that in the ergodic region [T>J 2(1 + rα)] the deviation of the total free energy of the Hopfield neural network converges in distribution asN → ∞ to a (shifted) Gaussian variable. Moreover, the free energy per site converges in probability to lim(1/N)ln≪ℤ N ≫.
Similar content being viewed by others
References
M. Aizenman, J. L. Lebowitz, and D. Ruelle,Commun. Math. Phys. 112:3–20 (1987).
D. Amit, M. Gutfreund, and A. Sampolinsky,Phys. Rev. A 35:2283 (1987);Ann. Phys. 173:30 (1987).
K. M. Kanin and Ya. G. Sinai,J. Stat. Phys. 20:573 (1979).
L. Pastur and M. Sherbina,J. Stat. Phys. 62:1 (1991).
S. Albeverio, B. Zegarlinski, and B. Tirozzi, Rigorous results for the free energy in the Hopfield model, Preprint.
J. L. Van Hemmen,Phys. Rev. A 34:3435–3445 (1986).
Author information
Authors and Affiliations
Rights and permissions
About this article
Cite this article
Scacciatelli, E., Tirozzi, B. Fluctuation of the free energy in the Hopfield model. J Stat Phys 67, 981–1008 (1992). https://doi.org/10.1007/BF01049007
Received:
Issue Date:
DOI: https://doi.org/10.1007/BF01049007