Balázs Sziklai (Corvinus, KRTK): Influence maximization on social networks: testing proxy-based algorithms

We devise a testing framework to rank proxy-based influence maximisation algorithms. Earlier works compare these algorithms by calculating their top choices on a given network, then checking which set fares better in a diffusion simulation. In real-life applications, however, the top choices of these algorithms might be unaccessible for various reasons. Consequently, we have to choose our spreaders from a different set that might not contain any highly ranked agents at all. There is no guarantee that a proxy that is better at predicting the performance of the most popular agents will be equally successful for an arbitrary group of individuals. This calls for a systematic test, and in this paper, we provide one with the help of a novel statistical method, the Sum of Ranking Differences. For demonstration, we use the real-life social network, iWiW and a classical diffusion model, the Linear Threshold. The final ranking of the proxies is remarkably different from what we obtain by examining the performance of their top choices. The results highlight that the standard test alone is an inadequate predictor of performance and SRD should be a necessary, if not the primary tool for ranking proxies

If you wish to receive a link for the zoom meeting on the day of the event, please send an email to Tamás Solymosi (tamas dot solymosi at uni dash corvinus dot hu)