Dear Students,

Tomorrow you will have the last two seminars of the Causal Network Course.

I and Dr. Justin Armanini will present you our research topics. Here a brief description of the content of the seminars:

Learning DBNs & Non-Homogeneous DBNs from Incomplete data in Intensive Care - Federico Pirola

Dynamic Bayesian networks (DBNs) are increasingly used in healthcare due to their ability to model complex temporal relationships. However, existing approaches to handling missing values fail to properly account for the temporal nature of the data and do not quantify uncertainty. In this seminar LUME-DBN and LUME-NH are presented as two novel full Bayesian methods to learn respectively homogeneous and non-homogeneous DBNs from Incomplete data sources. Two application in Intensive Care show the applicability of the methods in real-world observational scenarios.

A Gentle Introduction to Causal Mediation Analysis - Justin Armanini

In this seminar we will introduce the main concepts of causal mediation analysis from a theoretical perspective, clarifying how total, direct, and indirect effects can be defined and interpreted within a causal graph framework. We will then present an application in the clinical domain, using causal mediation to assess triage discrimination and to disentangle unfair from clinically justified disparities.

Tomorrow will also be the final lecture of the course, so we invite you to attend and ask any questions regarding the exam. Then, if you have any additional questions, please feel free to email me at: f.pirola17@campus.unimib.it

Best Wishes,
Federico Pirola