Uncertain Graph Processing through Representative Instances and Sparsification

Data in several applications can be represented as an uncertain graph, whose edges are labeled with a probability of existence. Currently, most query and mining tasks on uncertain graphs are based on Monte-Carlo sampling, which is rather time consuming for the large uncertain graphs commonly found in practice (e.g., social networks). To overcome the high cost, in this doctoral work we propose two approaches. The first extracts deterministic representative instances that capture structural properties of the uncertain graph. The query and mining tasks can then be efficiently processed using deterministic algorithms on these representatives. The second approach sparsifies the uncertain graph (i.e., reduces the number of its edges) and redistributes its probabilities, minimizing the information loss. Then, Monte-Carlo sampling applied to the reduced graph becomes much more efficient.