Abstract:An emerging trend in approximate counting is to show that certain `low-temperature' problems are easy on typical instances, despite worst-case hardness results. For the class of regular graphs one usually shows that expansion can be exploited algorithmically, and since random regular graphs are good expanders with high probability the problem is typically tractable. Inspired by approaches used in subexponential-time algorithms for Unique Games, we develop an approximation algorithm for the partition function of the ferromagnetic Potts model on graphs with a small-set expansion condition. In such graphs it may not suffice to explore the state space of the model close to ground states, and a novel feature of our method is to efficiently find a larger set of `pseudo-ground states' such that it is enough to explore the model around each pseudo-ground state.
What problem does this paper attempt to address?
The problem that this paper attempts to solve is to efficiently approximate the partition function of the ferromagnetic Potts model on a specific type of graph. Specifically, the author focuses on small - set expanders with small - set expansion conditions. In some cases, this type of graph may not satisfy the traditional expansion conditions, so new algorithms are required to handle these special cases.
### Main contributions of the paper
1. **New algorithm**: The paper proposes a new approximation algorithm for calculating the partition function of the ferromagnetic Potts model on graphs with small - set expansion conditions. This algorithm is particularly effective under low - temperature conditions (i.e., when β is large).
2. **Concept of pseudo - ground state**: Traditional methods usually rely on the ground state (i.e., a coloring where each edge has the same color), but in this type of graph, this method may not be applicable. The author introduces the concept of "pseudo - ground state" and approximates the partition function by finding a set of "pseudo - ground states".
3. **Application of spectral graph theory**: The author uses tools in spectral graph theory, especially properties related to the eigenvalues of the graph, to design and analyze the algorithm. This enables the algorithm to effectively handle graphs with small - set expansion conditions.
### Specific problem description
- **Ferromagnetic Potts model**: This is a probability distribution model defined on the vertex coloring of a graph, where each vertex can be assigned a color, and the same color of adjacent vertices will increase the energy of the system. The partition function \(Z_G(β)\) of the model is defined as:
\[
Z_G(β)=\sum_{\omega: V(G)\to[q]}e^{βm_G(\omega)}
\]
where \(m_G(\omega)\) represents the number of monochromatic edges in graph \(G\) under coloring \(\omega\), and \(β\) is the inverse temperature parameter.
- **Small - set expander graph**: The characteristic of this type of graph is that there are some small sets whose boundaries are relatively large compared to the set size. Specifically, for a graph \(G=(V, E)\), if for any set \(S\subset V\) satisfying \(|S|\leq|V|/2\), there is \(|\partial(S)|\geq\alpha|S|\), then \(G\) is called an \(\alpha\)-expander graph. Small - set expander graphs are a weakening of this condition.
### Solutions
- **Pseudo - ground state**: The author introduces the concept of "pseudo - ground state", that is, colorings that are not true ground states but are close to the ground state in some local regions. By finding these pseudo - ground states, the partition function can be more effectively approximated.
- **Polymer model**: The author uses the polymer model in statistical physics to transform the vertex coloring problem in the graph into a polymer distribution problem. Through the cluster expansion of the polymer model, the partition function can be approximately calculated.
- **Spectral graph theory**: Use tools in spectral graph theory, especially properties related to the eigenvalues of the graph, to design and analyze the algorithm. Specifically, the author uses the eigenvalues of the normalized Laplacian matrix of the graph to control the complexity and accuracy of the algorithm.
### Main results
- **Theorem 1**: For a \(\Delta\)-regular graph \(G\) with maximum degree \(\Delta\), if certain small - set expansion conditions are met, the partition function \(Z_G(β)\) of the ferromagnetic Potts model can be approximately calculated in polynomial time.
- **Theorem 2**: For a graph \(G\) with maximum degree \(\Delta\), if certain small - set expansion conditions are met, the partition function \(Z_G(β)\) of the ferromagnetic Potts model can be approximately calculated in polynomial time.
- **Theorem 3**: Provides a technical result for proving Theorem 1 and Theorem 2. This theorem describes how to divide a graph into several parts, each of which is an expander graph, and there are fewer connections between these parts.
### Applications and significance
- **Counting problem**: This algorithm can be applied to the approximate calculation of #BIS (counting independent set problem), especially.