Simultaneous influencing and mapping for health interventions

摘要: Influence Maximization is an active topic, but it was always assumed full knowledge of the social network graph. However, graph may actually be unknown beforehand. For example, when selecting a subset homeless population to attend interventions concerning health, we deal with that not fully known. Hence, introduce novel problem simultaneously influencing and mapping (i.e., learning) We study class algorithms, where show that: (i) traditional algorithms have arbitrarily low performance; (ii) can effectively influence map independence objectives hypothesis holds; (iii) does hold, upper bound for loss converges 0. run extensive experiments over four real-life networks, two alternative models, obtain significantly better results in both than approaches.