Speeding-up node influence computation for huge social networks

Abstract

We address the problem of efficiently estimating the influence degree for all the nodes simultaneously in the network under the SIR setting. The proposed approach is a further improvement over the existing work of the bond percolation process which was demonstrated to be very effective, i.e., three orders of magnitude faster than direct Monte Carlo simulation, in approximately solving the influence maximization problem. We introduce two pruning techniques which improve computational efficiency by an order of magnitude. This approach is generic and can be instantiated to any specific diffusion model. It does not require any approximations or assumptions to the model that were needed in We analyze how the performance changes when the network structure is systematically changed using synthetically generated networks and identify important factors that affect the performance.

Detecting Critical Links in Complex Network to Maintain Information Flow/Reachability

“…In order to overcome this problem, by borrowing and extending the basic ideas of pruning techniques proposed in [21,22], below we propose new acceleration techniques called marginal-link updating (MLU) and redundant-link skipping (RLS). The MLU technique locally updates the bottom-k sketches of some nodes when removing links incident to a node with in-degree 0 or out-degree 0 in the network G. First, let v ∈ V be a node with in-degree 0, i.e., |Q 1 (v; G)| = 0.…”

“…Unlike node-centrality measures derived only from network topology, influence degree centrality exploits a dynamical process on the network as well. An efficient method of simultaneously estimating the influence degrees of all the nodes was presented under the SIR model setting [22]. We note that influence degree centrality can also be employed for identifying super-mediators of information diffusion in the social network [28].…”

“…In addition, we utilized two synthetic networks with a DAG (Directed Acyclic Graph) property, which were generated by using the DCNN and DBA methods described in [21,22], respectively. For the sake of convenience, we briefly revisit these methods.…”