Some observations for mean-field spin glass models (Q2426697)

From MaRDI portal
Revision as of 21:03, 27 June 2024 by ReferenceBot (talk | contribs) (‎Changed an Item)
scientific article
Language Label Description Also known as
English
Some observations for mean-field spin glass models
scientific article

    Statements

    Some observations for mean-field spin glass models (English)
    0 references
    0 references
    0 references
    23 April 2008
    0 references
    The authors consider the question of universality, with respect to coupling distribution, of the infinite-volume free energy of some mean-field spin glass models, notably the Sherrington-Kirkpatrick (SK) model and the Viana-Bray (VB) one (the latter is a diluted model, i.e., the couplings form a sparse random graph). For SK, S. Starr and B. Vermesi prove that the free energy depends only on the first two moments of the couplings, thereby improving a previous result by P. Carmona and Y. Hu, who additionally required finiteness of the couplings' third moment. An analogous universality result for VB is contained in Corollary 1.5. In section 2, for both models it is proven that the infinite-volume free energy is unchanged if a (very mild) form of statistical dependence among couplings is introduced. For instance, this is the case for SK if the couplings \(J_{ij}\), instead of being i.i.d. standard Gaussian, are such that the vector \(\{J_{ij}\}_{1\leq i,j\leq N}\) is uniformly distributed on the unit sphere \(S^{N^2-1}\). For VB, the authors prove the intuitive fact that fixing the number of couplings to \(cN\) (\(c\) a positive constant) or letting it be a Poisson random variable with average \(cN\) is ininfluent on the thermodynamic limit.
    0 references
    spin glasses
    0 references
    Viana-Bray model
    0 references
    Sherrington-Kirkpatrick model
    0 references
    infinitely divisible distributions
    0 references

    Identifiers

    0 references
    0 references
    0 references
    0 references
    0 references
    0 references