Two-distance minimal dominating set problem studied by statistical mechanics and simulated annealing

TL;DR

This paper applies spin glass theory and simulated annealing to analyze the 2-distance minimal dominating set problem, revealing phase transition behaviors and improving solution estimation methods on random networks.

Contribution

It extends the cavity method to the 2-distance MDS problem, analyzing entropy and convergence, and compares belief propagation with other algorithms for better solutions.

Findings

Belief propagation does not converge above a certain temperature threshold.

Entropy density is zero at finite inverse temperature for specific degree ranges.

Belief propagation decimation outperforms greedy heuristics.

Abstract

The L-distance (especially the 2-distance) minimal dominating set (MDS) problem is widely considered in various dominating set problems. Recently, we studied the regular dominating set problem using the cavity method and developed two algorithms (belief propagation decimation and survey propagation decimation) to estimate the solution of a given graph, resulting in very good estimations of the minimal dominating size. This paper describes the development of spin glass theory to study the 2-distance MDS problem. First, we show that the belief propagation equation does not converge when the inverse temperature $\beta$ is greater than a certain threshold value on both regular and Erdos--Renyi random networks. Second, we find that the entropy density is equal to zero at finite inverse temperature on regular random graphs when the node degree is from 3--9, and on Erdos--Renyi random networks when the node degree is from 4.2--10.4; the entropy density is positive in all other cases. This result is proved using a dynamical simulated annealing process. Third, the results of replica symmetry theory are shown to be in agreement with those of the belief propagation algorithm, and the results of the belief propagation decimation algorithm are found to be better than those of the greedy heuristic algorithm.