## Seminaire LIA: Yezekael Hayel (LIA)

Yezekael Hayel: 15 Mars 2018, à 11h00 salle 6.

Titre: Opinion dynamics in large social networks based on epidemics on graphs

Abstract: Understanding propagation, spreading and epidemics over graphs have several many application domains (industrial engineering, networking, biology, social networks, ...). In this talk we consider a stochastic propagation model in order to study opinion dynamics of agents in large social networks. We are interested in a multi-leveled opinion of each agent which is randomly influenced by the binary actions of its neighbors. It is shown that, as far as the number of agents in the network is finite, the model asymptotically produces consensus. The consensus value corresponds to one of the absorbing states of the associated Markov system. However, when the number of agents is large, we emphasize that partial agreements are reached and these transient states are metastable, i.e., the expected persistence duration is arbitrarily large. These states are characterized using an N-intertwined mean field approximation (NIMFA) for the Markov system modeling the opinion dynamics. Numerical simulations validate the proposed analysis.