Experiments with Declarative Modeling of Maximum Clique Problem Using Solvers Supported by MiniZinc

In this paper, we present the results of solving the Maximum Clique Problem using declarative modeling approaches. Our goal is to create a single reusable model that can be utilized to compare several solvers supported by MiniZinc. For the comparison, we have used data sets from the DIMACS benchmarks. We present the performance results of the solvers used in experiments using single as well as 16-threaded configurations on data set instances of 125, 250, 500, 1000, and 2000 nodes. Our initial results revealed that the Gurobi solver handles the small data sets better than the Gecode or COIN-OR, but the Gecode solver is better over the larger data sets. We found that COIN-OR was not able to find a solution for data sets of more than 500 nodes, on the single thread configuration, within our chosen time limit of 120s.