Influence model

Cascade Models. We focus on two widely adopted diffu- sion models: the Independent Cascade (IC) model and Linear Threshold (LT) model. Both models share the same discrete- time cascade process as mentioned in Section 2.1 and the main difference lies in how the inactive nodes get activated:

• IC model. Suppose node u gets activated at timestamp i, then u has a single chance to activate its inactive out- neighbor v with probability p(u,v) at timestamp i + 1.

• LT model. In the LT model, it assumes that for each node v: (i) the sum of the propagation probability of its incom- ing edges is no more than 1, and (ii) a probability λv is selected uniformly at random from [0, 1]. If v is inactive at timestamp i, then it becomes activated at timestamp i + 1 if and only if Íu∈A p(u,v) ≥ λv, where A is the set of activated in-neighbor of v at timestamp i.

> Influence Maximization Revisited: Efficient Reverse Reachable Set Generation with Bound Tightened