WorksComputer Science
Multi-seed information diffusion with varying reproduction
Abstract
We define the -process that generalizes the notion of Hamiltonicity through the lense of information diffusion. We say that a graph is -infectable if it can be fully infected by transmission trees seeded by at most initial infections in which each infection leads to at most others; or, equivalently, if it can be exactly vertex-covered by at most trees of degree at most . While determining infectability is NP-hard in general, we argue fixed-parameter tractability when parameterised by treewidth. Additionally, we give computational experimental results comparing the performance of several natural heuristics for minimising the number of initial infections required to an optimal LP-based approach on several standard complex network models. We find that a uniformly random approach outperforms greedy heuristics based on degree or clustering.