Cooperative Game Theoretic Solution Concepts for top-$k$ Problems

The problem of finding the k most critical nodes, referred to as the top-k problem, is a very important one in several contexts such as information diffusion and preference aggregation in social networks, clustering of data points, etc. It has been observed in the literature that the value allotted to a node by most of the popular cooperative game theoretic solution concepts, acts as a good measure of appropriateness of that node (or a data point) to be included in the top-k set, by itself. However, in general, nodes having the highest k values are not the desirable top-k nodes, because the appropriateness of a node to be a part of the top-k set depends on other nodes in the set. As this is not explicitly captured by cooperative game theoretic solution concepts, it is necessary to post-process the obtained values in order to output the suitable top-k nodes. In this paper, we propose several such post-processing methods and give reasoning behind each of them, and also propose a standalone algorithm that combines cooperative game theoretic solution concepts with the popular greedy hill-climbing algorithm.