Searcharxiv⌕ Search

arXiv subjects

Anton Bovier

Publications and source records attributed to Anton Bovier.

67 records · Page 4Linked to original sources

The spin-glass phase-transition in the Hopfield model with p-spin interactions

We study the Hopfield model with pure $p$-spin interactions with even $p\geq 4$, and a number of patterns, M(N) growing with the system size, $N$, as $M(N) = \a N^{p-1}$. We prove the existence of a critical temperature $\b_p$ characterized as the first time quenched and annealed free energy differ. We prove that as $p\uparrow\infty$, $\b_p\to\sqrt {\a 2\ln 2}$. Moreover, we show that for any $\a>0$ and for all inverse temperatures $\b$, the free energy converges to that of the REM at inverse temperature $\b/\sqrt\a$. Moreover, above the critical temperature the distribution of the replica overlap is concentrated at zero. We show that for large enough $\a$, there exists a non-empty interval of in the low temperature regime where the distribution has mass both near zero and near $\pm 1$. As was first shown by M. Talagrand in the case of the $p$-spin SK model, this implies the the Gibbs measure at low temperatures is concentrated, asymptotically for large $N$, on a countable union of disjoint sets, no finite subset of which has full mass. Finally, we show that there is $\a_p\sim 1/p!$ such that for $\a>\a_p$ the set carrying almost all mass does not contain the original patterns. In this sense we describe a genuine spin glass transition. Our approach follows that of Talagrand's analysis of the $p$-spin SK-model. The more complex structure of the random interactions necessitates, however, considerable technical modifications. In particular, various results that follow easily in the Gaussian case from integration by parts fromulas have to be derived by expansion techniques.

cond-mat.dis-nn↗

Metastability in Glauber dynamics in the low-temperature limit: Beyond exponential asymptotics

We consider Glauber dynamics of classical spin systems of Ising type in the limit when the temperature tends to zero in finite volume. We show that information on the structure of the most profound minima and the connecting saddle points of the Hamiltonian can be translated into sharp estimates on the distribution of the times of metastable transitions between such minima as well as the low lying spectrum of the generator. In contrast with earlier results on such problems, where only the asymptotics of the exponential rates is obtained, we compute the precise pre-factors up to multiplicative errors that tend to 1 as $T\downarrow 0$. As an example we treat the nearest neighbor Ising model on the 2 and 3 dimensional square lattice. Our results improve considerably earlier estimates obtained by Neves-Schonmann and Ben Arous-Cerf. Our results employ the methods introduced by Bovier, Eckhoff, Gayrard, and Klein.

cond-mat.dis-nn↗

Sample path large deviations for a class of Markov chains related to disordered mean field models

We prove a large deviation principle on path space for a class of discrete time Markov processes whose state space is the intersection of a regular domain $Ł\subset \R^d$ with some lattice of spacing $\e$. Transitions from $x$ to $y$ are allowed if $\e^{-1}(x-y)\in \D$ for some fixed set of vectors $\D$. The transition probabilities $p_\e(t,x,y)$, which themselves depend on $\e$, are allowed to depend on the starting point $x$ and the time $t$ in a sufficiently regular way, except near the boundaries, where some singular behaviour is allowed. The rate function is identified as an action functional which is given as the integral of a Lagrange function. %of time dependent relativistic classical mechanics. Markov processes of this type arise in the study of mean field dynamics of disordered mean field models.

math.PR↗

Stochastic symmetry-breaking in a gaussian Hopfield model

We study a ``two-pattern'' Hopfield model with Gaussian disorder. We find that there are infinitely many pure states at low temperatures in this model, and we find that the metastate is supported on an infinity of symmetric pairs of pure states. The origin of this phenomenon is the random breaking of a rotation symmetry of the distribution of the disorder

cond-mat.dis-nn↗

Comment on ``Capacity of the Hopfield model''

In a recent paper ``The capacity of the Hopfield model, J. Feng and B. Tirozzi claim to prove rigorous results on the storage capacity that are in conflict with the predictions of the replica approach. We show that their results are in error and that their approach, even when the worst mistakes are corrected, is not giving any mathematically rigorous results.

cond-mat.dis-nn↗

Metastates in the Hopfield model in the replica symmetric regime

We study the finite dimensional marginals of the Gibbs measure in the Hopfield model at low temperature when the number of patterns, $M$, is proportional to the volume with a sufficiently small proportionality constant $\a>0$. It is shown that even when a single pattern is selected (by a magnetic field or by conditioning), the marginals do not converge almost surely, but only in law. The corresponding limiting law is constructed explicitly. We fit our result in the recently proposed language of ``metastates'' which we discuss in some length. As a byproduct, in a certain regime of the parameters $\a$ and $\b$ (the inverse temperature), we also give a simple proof of Talagrand's [T1] recent result that the replica symmetric solution found by Amit, Gutfreund, and Sompolinsky [AGS] can be rigorously justified.

cond-mat.dis-nn↗

Hopfield models as generalized random mean field models

We give a comprehensive self-contained review on the rigorous analysis of the thermodynamics of a class of random spin systems of mean field type whose most prominent example is the Hopfield model. We focus on the low temperature phase and the analysis of the Gibbs measures with large deviation techniques. There is a very detailed and complete picture in the regime of ``small $\a$''; a particularly satisfactory result concerns a non-trivial regime of parameters in which we prove 1) the convergence of the local ``mean fields'' to gaussian random variables with constant variance and random mean; the random means are from site to site independent gaussians themselves; 2) ``propagation of chaos'', i.e. factorization of the extremal infinite volume Gibbs measures, and 3) the correctness of the ``replica symmetric solution'' of Amit, Gutfreund and Sompolinsky [AGS]. This last result was first proven by M. Talagrand [T4], using different techniques.

cond-mat.dis-nn↗

The low-temperature phase of Kac-Ising models

We analyse the low temperature phase of ferromagnetic Kac-Ising models in dimensions $d\geq 2$. We show that if the range of interactions is $\g^{-1}$, then two disjoint translation invariant Gibbs states exist, if the inverse temperature $\b$ satisfies $\b -1\geq \g^\k$ where $\k=\frac {d(1-\e)}{(2d+1)(d+1)}$, for any $\e>0$. The prove involves the blocking procedure usual for Kac models and also a contour representation for the resulting long-range (almost) continuous spin system which is suitable for the use of a variant of the Peierls argument.

cond-mat↗

Distribution of overlap profiles in the one-dimensional Kac-Hopfield model

We study a one-dimensional version of the Hopfield model with long, but finite range interactions below the critical temperature. In the thermodynamic limit we obtain large deviation estimates for the distribution of the ``local'' overlaps, the range of the interaction, $γ^{-1}$, being the large parameter. We show in particular that the local overlaps in a typical Gibbs configuration are constant and equal to one of the mean-field equilibrium values on a scale $o(\g^{-2})$. We also give estimates on the size of typical ``jumps''. i.e. the regions where transitions from one equilibrium value to another take place. Contrary to the situation in the ferromagnetic Kac-model, the structure of the profiles is found to be governed by the quenched disorder rather than by entropy.

cond-mat↗

Schrödinger operators generated by substitutions

Schrödinger operators with potentials generated by primitive substitutions are simple models for one dimensional quasi-crystals. We review recent results on their spectral properties. These include in particular an algorithmically verifiable sufficient condition for their spectrum to be singular continuous and supported on a Cantor set of zero Lebesgue measure. Applications to specific examples are discussed.

cond-mat↗

There are No Nice Interfaces in 2+1 Dimensional SOS-Models in Random Media

We prove that in dimension $d\leq 2$ translation covariant Gibbs states describing rigid interfaces in a disordered solid-on-solid (SOS) cannot exist for any value of the temperature, in contrast to the situation in $d\geq 3$. The prove relies on an adaptation of a theorem of Aizenman and Wehr. Keywords: Disordered systems, interfaces, SOS-model

cond-mat↗

The Retrieval Phase of the Hopfield Model: A Rigorous Analysis of the Overlap Distribution

Standard large deviation estimates or the use of the Hubbard-Stratonovich transformation reduce the analysis of the distribution of the overlap parameters essentially to that of an explicitly known random function $Φ_{N,\b}$ on $\R^M$. In this article we present a rather careful study of the structure of the minima of this random function related to the retrieval of the stored patterns. We denote by $m^*(\b)$ the modulus of the spontaneous magnetization in the Curie-Weiss model and by $\a$ the ratio between the number of the stored patterns and the system size. We show that there exist strictly positive numbers $0<\g_a<\g_c$ such that 1) If $\sqrt\a\leq \g_a (m^*(\b))^2$, then the absolute minima of $Φ$ are located within small balls around the points $\pm m^*e^μ$, where $e^μ$ denotes the $μ$-th unit vector while 2) if $\sqrt\a\leq \g_c (m^*(\b))^2$ at least a local minimum surrounded by extensive energy barriers exists near these points. The random location of these minima is given within precise bounds. These are used to prove sharp estimates on the support of the Gibbs measures. KEYWORDS: Hopfield model, neural networks, storage capacity, Gibbs measures, self-averaging, random matrices

cond-mat↗

An almost sure large deviation principle for the Hopfield model

We prove a large deviation principle for the finite dimensional marginals of the Gibbs distribution of the macroscopic `overlap'-parameters in the Hopfield model in the case where the number of random patterns, $M$, as a function of the system size $N$ satisfies $\limsup M(N)/N=0$. In this case the rate function (or free energy as a function of the overlap parameters) is independent of the disorder for almost all realization of the patterns and given by an explicit variational formula.

cond-mat↗