Structure and Dynamics of Information Pathways in On-line Media
Diffusion of information, influence and diseases are stochastic processes that occur over highly dynamic
networks. We observe an endless flow of information that propagates over a rapidly time-varying network. Unfortunately the network
is often hidden or unobserved, but key to understand and predict information propagation.
We have developed INFOPATH, an on-line algorithm that relies on stochastic convex optimization to find the time-varying networks over
which information propagates from the temporal traces of diffusion.
The YouTube video on the left shows a few insights from our results. Websites are
represented by circles. Bigger circles corresponds to websites that publish more
relevant information about a topic. Sites are automatically divided into communities, and
each community is represented by a different color. A link (edge) from a website A to a website B indicates
that B tends to mention the same information content just after A..