1: \begin{abstract}

2: %Counting is one of the fundamental problems in combinatorial theory, whose  importance has been emphasized by Valiant's

3: %$\#P$ hardness  class.  While problems in $\#P$ are unlikely to be solvable exactly, some of the counting problems

4: %can be solved approximately using  powerful sampling techniques based on the theory of rapidly mixing Markov chains.

5: 

6: We propose a  new type of approximate counting algorithms for the

7: problems of enumerating the number of independent sets and proper

8: colorings in low degree graphs with large girth. Our algorithms are

9: not based on a commonly used Markov chain technique, but rather are

10: inspired by developments in statistical physics in connection with

11: correlation decay properties of Gibbs measures and its implications

12: to uniqueness of Gibbs measures on infinite trees, reconstruction

13: problems and local weak convergence methods.

14: 

15: On a negative side, our algorithms provide $\epsilon$-approximations

16: only to the logarithms of the size of a feasible set (also known as

17: free energy in statistical physics). But on the positive side, our

18: approach provides deterministic as opposed to probabilistic

19: guarantee on approximations. Moreover, for some regular graphs we

20: obtain explicit values for the counting problem. For example, we

21: show that every $4$-regular $n$-node graph with large girth has

22: approximately $(1.494\ldots)^n$ independent sets, and in every

23: $r$-regular graph with $n$ nodes and large girth the number of

24: $q\geq r+1$-proper colorings is approximately $[q(1-{1\over

25: q})^{r\over 2}]^n$, for large $n$. In statistical physics

26: terminology, we compute explicitly the limit of the log-partition

27: function. We extend our results to random regular graphs. Our

28: explicit results would be hard to derive via the Markov chain

29: method.

30: \end{abstract}

31: