9
$\begingroup$

For non-Ising'ers: Given a graph, we study the probability-distribution on the set of colorings ("Spin-up" and "-down") generated by a given correlation ("force to equality") between adjacient nodes (Bolzmann-Term depending exponentially on a "Temperature Parameter"). The limit of high Temperature is no-correlation whatsoever, the limit of low Temerature is all colors coincide for sure (are "parallel").

In this paper (Dommers, Giardinà, van der Hofstad, 2010) the partition function of the Ising model on a sparse random graph was calculated (sparse = almost surely no triangle, quadrangle etc )...well, I must add "calculated" refers to a random-variable fixpoint equation - I've put some effort into deriving an explicit expression thereof, but did not suceed ... a different story and maybe later a different question ;-)

The proof techniques very basically relay on the transfer matrix method for a local tree inside the random graph (which is defined by a given branching distribution, e.g. Poisson); this ultimately leads to a recursion formula for the up/down-random-variable of a single site and the authors proof the existence of a unique solution (the mentionen fix-point). Partition function etc. are derived as various expectation vals.

For use in an own work I would like to get ahold of the correlators, i.e. the probability of two sites of given distance $a$ to coincide. Though I do not expect this to be easier than explicitely calculate the partition function, a similar fix-point equation as in the original work would totally suffice! I tried the trick one uses in 1D, but I didn't get through...although I would naively expect, that it's easier than 2D, as you have better control over the walks...?

Moreover, it would be interesting to have multi-correlators in the sense of knowing the probability of given up-down configurations e.g. in an arbitrary $abc$-triangle depending on the edge-length's $a,b,c$. $\;\;$ I don't even recall such from classical (2D) Ising model....any sources or ideas? Is that in "physically-sound" language the n-point-functions in this case?

$\endgroup$

1 Answer 1

4
$\begingroup$

There is a method to study these physical systems called belief propagation (BP), which yields exact methods for interaction graphs that are trees, and empirically works pretty fine when you have physical systems with loopy interaction graphs that locally-look-like trees. Randomly-generated sparse interaction graphs are of this form (its explicitly mentioned in the paper), and you could apply Belief Propagation to your model to obtain good estimates of the following quantities.

  • Marginal probabilities $\sigma_i$.
  • Joint probabilities $p(\sigma_i, \sigma_j)$, in particular, the correlators you mention.
  • Conditional probabilities $p(\sigma_i | \sigma_j)$
  • Partition functions, energies, free energies, entropies.

And, in addition, BP can be adapted to sample from the probability distribution $p(\sigma)$. In physics, systems like these are sometimes called 'diluted spin-glasses'.

I summarise the mean features of Belief propagation below. Mézard, Montanari's book contains a very pedagogical exposition of the method, also rather up-to-date and complete; you can read the draft of the book online without buying it. I personally used this book to prepare a seminar and I can recommend it. The connections between Belief Propagation and the ferromagnetic spin Isin chain are explained in Part D, section 14.1 of the online version.

Overview

Belief propagation is an iterative "message-passing" algorithm. For each site/spin/particle (vertexes of the graph) of your system, the belief propagation methods assigns incoming and leaving "messages" to its neighbouring edges (which represent interactions) in the interaction graph. These messages can be understood as "probability weights" that contribute to the marginal probability that the particle say on site $i$ is on a given local configuration $\sigma_i$, which I denote by $p(\sigma_i)$. The messages are updated through local computations (thus, not costly) for each particle, following a set of equations known as the belief-propagation update-rules. I will not write the update-rules here, for they are lengthy.

What it is important is the following: if on site $i$, with neighbours indexed by $a$, we denote the leaving/incoming messages on time $t$ by $p_{i\rightarrow a}^{t}(\sigma_i)$ and $p_{a\rightarrow i}^{t}(\sigma_i)$, then BP method gives a way to estimate the marginal probability $p(\sigma_i)$ as follows: $$p(\sigma_i)\simeq p^{t}(\sigma_i)=\prod_{a\in\partial_i}p_{a\rightarrow i}^{t-1}(\sigma_i)$$

For trees, this iteration converges to fix point in finite time $t_{*}$, and this bound has a known value. Moreover, the method provably computes the marginal probabilities exactly. In fact, for trees this method and transfer-matrix methods used e.g. for the classical Ising spin chain turn out to be equivalent.

If your system it is not a tree, you do not get exact results and the message-update iteration is not promised to converge. However, the method has been used with wide experimental success on graphs that locally-look-like-trees (this seems to set an upper bound to the correlation-distance that you can have) to provide good estimations of these marginal probabilities.

Finally, the method can be adapted to estimate all the other quantities I mentioned.

Drawbacks

The main disadvantage of this method is that it is rather complicated to find analytical bounds for errors or analytical conditions that guarantee convergence; both are fields of current research. For a recent study regarding convergence you can check [1]; regarding errors, I found this ones [2].

And, of course, another disadvantage is that if you add a lot of random interactions to your graph, at some point your interaction-graph stops being tree-like and the method would behave poorly. In fact, there is a threshold were a gigantic connected component appears in the interaction graph [3].

$\endgroup$
4
  • 1
    $\begingroup$ This is answer is very useful for me. Let me mention that Belief Propagation is practical algorithm for decoding error-correction codes like LDPC codes. I am wondering how this is connected with the Ising model ? May be I will ask it as a separate question later. $\endgroup$ Apr 23, 2012 at 5:45
  • 1
    $\begingroup$ The connection between the classical ferromagnetic Ising chain (Ising model on a line) and BP is one of the first examples worked out in the book I mentioned. Follow the link and check chapter D, section 14.1. Briefly, the interaction-graph of the model defines naturally a factor graph for BP. The messages are obtained re-writing the formulas for the marginals in a different way, using the Boltzmann distribution. The book also covers the connections with LDPC codes. $\endgroup$ Apr 23, 2012 at 11:45
  • 1
    $\begingroup$ By the way, the drawbacks I mentioned can become serious issues depending of your application. Since these are research-topic themselves, if you want to know more about errors/bounds, we would open a new thread, maybe in Theoretical Physics Stack Exchange. $\endgroup$ Apr 23, 2012 at 11:50
  • $\begingroup$ If you might comment on that question: mathoverflow.net/questions/94931/… I would be indebted $\endgroup$ Apr 23, 2012 at 12:16

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.