Algorithms for alpha-rate domination problems on weighted graphs

In this article, we investigate a domination set problem variant on vertex-weighted graphs. In the last few years, several algorithms have been presented for solving the minimum alpha and alpha-rate domination problem (also known as the positive influence dominating sets problem) on simple graphs. We recently proposed an algorithm for alpha-rate domination on weighted graphs based on randomised rounding of the solution of a linear programming formulation of this problem. Due to the use of linear programming, such an algorithm could be relatively time consuming for larger graphs. Here, we propose a new version using the divide and conquer technique, which uses a graph's community structure to create a solution from the solutions obtained on denser subgraphs (with some adjustments, if necessary). We also investigate greedy techniques for this problem using three different initial vertex selection strategies. We compare two different randomised rounding and three greedy algorithms on three different families of randomly generated graphs and on four real-world graphs obtained from a Twitter mentions network. Our results show that on dense random graphs the divide and conquer technique produces results comparable in total weight to the unembellished randomised rounding method, but signifficantly faster. For graphs with intrinsic modular structure, the divide and conquer technique actually produces better results. When the running time is prioritised over the optimality of results, two of the three explored greedy algorithms strategies perform better than the simple strategy of picking always vertices with the smallest weights. Also, greedy techniques outperform randomised algorithms on the very sparse Twitter graphs, and on the random dense graphs for high thresholds.