\paragraph{Related works.}
The study of edge prediction in graphs has been an active field of research for
over a decade~\cite{Nowell08, Leskovec07, AdarA05}. MLE estimation (regularized
and un-regularized) for Graph Inference has been studied both for discrete-time
models \cite{Netrapalli:2012, pouget} and continous-time models
\cite{GomezRodriguez:2010, gomezbalduzzi:2011, Abrahao:13}

More recently, the continuous-time processes studied in previous work have been
reformulated as a Hawkes processes, with recent papers
\cite{linderman2014discovering, linderman2015scalable, simma2012modeling},
focusing on Expectation-Maximization, Gibbs sampling and Variational Inference
methods. In comparison the discrete-time nature of the GLC model allows us to
scale the inference methods to larger graphs. Furthermore, while the works on
Hawkes processes exclusively used factorized graph priors, we hope that
Bayesian Inference for the GLC model will be able to accommodate more
expressive graph priors more easily. This is a direction we wish to explore in
future works.

The Active Learning formulation is, to the best of the authors' knowledge,
novel in this context. The Active Learning approach of \cite{shababo} share
some similarities with ours even though their model is not, strictly speaking,
a cascade model (in particular, the time steps are independent).