Log Gaussian Cox Process, Cross Validation, Information Criteria and Variational Bayes

Speaker: Hans A. Montcho, KAUST

Abstract: In this talk, I will summarize some ideas that I have explored during my PhD. The central topic is the log Gaussian Cox process (LGCP), one of the most used point process models. In the first half, I will discuss the notion of cross validation and information criteria for point process models as part of assessment tools for the LGCP. In the second half, I will discuss a fully deterministic variational Bayes strategy to go beyond Gaussian approximation for the class of Latent Gaussian Models. This latest work not only enhances the accuracy of the INLA method but also paves the way for next-generation latent non-Gaussian models.

Biography: Hans Montcho is a PhD candidate in statistics working under Prof Havard Rue and Prof David Bolin in the BayesComp group at KAUST University. His thesis is centered around Latent Gaussian Models, Bayesian computation, INLA, and model assessment with a particular attention to Point process models.

Host: Prof. A. Mira