Skip to Main Content

Graduate Student Seminar

Wednesday, April 18, 2018
11:00 AM - 12:00 PM
Fine Arts : 215
Session ChairGuy Djokam
DiscussantDr. Kathleen Hoffman

Speaker 1: Michael   Retzlaff
Optimization Based Domain Decomposition for Partial Differential Equations
We discuss a strategy for finding the numerical solution of a PDE using a formulation as a constrained minimization problem, first proposed by Gunzburger et al (1999).  The method allows for uncoupled PDE solves within each subdomain, so the bulk of the computations are of smaller scale and parallelizable.  Then a system of equations is constructed to find a set of control variables, which are the normal derivatives of the solution at the subdomain boundaries.  This system is in the form of a regularized least squares problem and is global in nature, so we strive to develop strategies for solving the potentially large system in an efficient manner.  After a discussion of the method, we will describe a 2-dimensional implementation using finite elements, present some numerical results of the implementation, and discuss our efforts to find efficient solvers for the control system.

Speaker 2: Michael Lucagbo
Effects of Education on Climate Risk Vulnerability in the Philippines: Evidence from Panel Data
The effects of climate change are being felt among countries with high vulnerability and low resilience to climate change. This study assesses the effect of increasing the level of education in improving resilience in the Philippines. The data form a combined time-series cross-section panel. The fixed-effects Poisson (FEP) regression model developed by Hausman et al (1984), which enables researchers to model count behavior in the context of panel data, is adopted. The likelihood function is globally concave so maximum likelihood estimation is used. Moreover, because of the presence of over dispersion, White's heteroskedasticity-consistent standard errors are used as correction. The results show significant effect of education in reducing number of deaths by natural disasters.