Quantification of uncertainties in models and data constitutes the fundamental basis for quantifying risk, thus being able to make informed engineering decisions. Towards this, uncertainties associated with chronic deteriorating stressors (e.g. corrosion or fatigue) as well as with recurrent hazards (e.g. earthquakes or hurricanes) need to be efficiently processed, either directly, based on data, or indirectly, propagated through physics-based models. The goal is to efficiently estimate and infer probabilistic quantities of interest, but also to learn statistical structures that are able to forecast the dynamics of these quantities in the absence of physics-based simulators. Probabilistic modeling in this regard has to accommodate the presence of high dimensional feature and random variable spaces, as well as the presence of correlations and dependencies at the spatial and temporal scales. This line of research is involved in the study and development of such risk & reliability methods as these pertain to probabilistic performance-based engineering, structural fragility analysis, Bayesian networks inference and learning, and supervised or unsupervised probabilistic machine learning. 

​