No abstract

In a previous work, the n-vicinity method for approximate calculation of the partition function of a spin system was proposed. The equation of state was obtained in the most general form. In the present paper, we analyze the applicability of this method for the Ising model on a D-dimensional cubic lattice. The equation of state is solved for an arbitrary dimension D and the behavior of the free energy is analyzed. As expected, for large dimensions, 3 D  , the system demonstrates a phase transition of the second kind. In this case, we obtain an analytical expression for the critical value of the inverse temperature. When 37 D  this expression is in a very good agreement with the results of computer simulations. In the case of small dimensions 3 D  , there is a noticeable discrepancy with the known exact results.

Abstarct. We examined energy spectrums of some particular systems of N binary spins. It is shown that the configuration space can be divided into N classes, and in the limit N   the energy distributions in these classes can be approximated by the normal distributions. For each class we obtained the expressions for the first three moments of the energy distribution, including the case of presence of a nonzero inhomogeneous magnetic field. We also derived the expression for the variance of the quasienergy distribution in the local minimum. We present the results of computer simulations for the standard Ising model and the Sherrington-Kirkpatrick and Edwards-Anderson models of spin glass. Basing on these results, we justified the new method of the partition function calculation.

We examine connection matrices of Ising systems with long-rang interaction on d-dimensional hypercube lattices of linear dimensions L. We express the eigenvectors of these matrices as the Kronecker products of the eigenvectors for the one-dimensional Ising system. The eigenvalues of the connection matrices are polynomials of the dth degree of the eigenvalues for the one-dimensional system. We show that including of the long-range interaction does not remove the degeneracy of the eigenvalues of the connection matrix. We analyze the eigenvalue spectral density in the limit L → ∞. In the case of the continuous spectrum, for d ⩽ 2 we obtain analytical formulas that describe the influence of the long-range interaction on the spectral density and the crucial changes of the spectrum.

We generalize the standard Hopfield model to the case when a weight is assigned to each input pattern. The weight can be interpreted as the frequency of the pattern occurrence at the input of the network. In the framework of the statistical physics approach we obtain the saddle-point equation allowing us to examine the memory of the network. In the case of unequal weights our model does not lead to the catastrophic destruction of the memory due to its overfilling (that is typical for the standard Hopfield model). The real memory consists only of the patterns with weights exceeding a critical value that is determined by the weights distribution. We obtain the algorithm allowing us to find this critical value for an arbitrary distribution of the weights, and analyze in detail some particular weights distributions. It is shown that the memory decreases as compared to the case of the standard Hopfield model. However, in our model the network can learn online without the catastrophic destruction of the memory.