Marked temporal point processes (MTPP) have emerged as a powerful modeling framework to capture the underlying 
generative mechanism of asynchronous events localized in continuous time. Recent MTPP methods use deep 
architectures to approximate the associated conditional intensity function that in turn models the entire 
point process. However, these methods apply certain constraints to the model design and depend largely on the 
input data for their predictions. Our recent work uses TPPs for identifying latent and overlapping communities 
in geo-tagged networks with significant improvement over the previous state-of-the-art methods. We learn a 
multivariate Hawkes process with each process for an individual user in the network. We also present another 
model that provides a novel modeling framework for MTPPs in presence of missing events. More specifically, 
we first model the generative processes of observed events and missing events using two MTPPs and jointly learn 
both the MTPPs by variational inference.