Abstract for Prof. Eleftheria Kontou's talk
We investigate the problem of evacuating residents within an urban network where we have vehicles of different fuel types. While conventional fossil-fueled vehicles typically have access to a well-established network of refueling stations, vehicles of newer and alternative fuel technologies still face limitations due to sparse refueling and charging station networks and, consequently, have difficulties accessing refueling infrastructure reliably. Hence, during an evacuation procedure, an adversary may decide to pose a hybrid threat. While the immediate natural or anthropogenic threat (e.g., hurricane, nuclear plant accident, flooding, etc.) is taking place, an adversary may opt to “attack” the evacuation plan to cause the evacuation to fail. The adversarial attack can be physical (as in, attacking existing infrastructure) or virtual (in the form of a cyber-attack or a misinformation campaign). Hence, it becomes important for evacuation planners and managers to know in advance the roads that are fundamental to the plan’s success to fortify them and consider alternatives in case of failure during the evacuation. The novelty in our approach lies in the fact that we consider different vehicles with different refueling needs when making the decisions for fortifying part of the evacuation plan.
Abstract for Wendy Tam Cho's talk
The United States is in a state of crisis stemming from police abuse, racial injustice, and pandemic unrest—factors that are not only concurrent, but likely overlapping. The convergence of these events lays bare long standing societal issues and provides us with an opportunity, indeed, behooves us to examine the role of structural racism in population health. We utilize the Electronic Health Records from the UCSF Health system to examine the underlying biomedical reasons for observed racial health disparities and the relationship between these disparities and parallel inequalities in other sectors. We conduct standard statistical analyses as well as enhance causal inference models for these observational data by improving the computational capabilities of these models so that they are able to scale to large size and feature-rich digitized health records.