Séminaire des équipes de recherche
The effect of network topology on human and machine communications
Kevin Scaman from Microsoft Research-Inria Joint Center (Paris-Seattle) will be speaking about "the effect of network topology on human and machine communications".
- Date : 24/03/2017
- Lieu : Inria Paris - 11h00 - Salle C334
- Intervenant(s) : Kevin Scaman (Microsoft Research-Inria)
How fast does information propagate in a network? And how can we leverage this diffusion for marketing or optimization purposes? To answer these questions, I will investigate two case scenarios: human to human communications on social networks and machine to machine communications for distributed optimization. In the former, I will show that the spectral radius, a quantity related to a network's connectivity, plays a large role in the explosiveness of information cascades. In the latter, I will provide optimal convergence rates for the distributed optimization of strongly convex and smooth functions when machines are bound to communicate across a network of limited bandwidth.
Both examples will highlight the role of spectral characteristics of networks as a natural criterion to assess the connectivity of a network and its ability to spread information efficiently.