VFF: Data+ student projects: Gerrymandering & Geometry of Weather

This week we will hear from two teams of students who participated in the Data+ program with the Information Initiative at Duke (iiD) over the summer. "Gerrymandering": Our team set out to develop a method to quantify the effects of gerrymandering in the United States on electoral outcomes. We used a Markov Chain Monte Carlo simulation to produce several samples of possible districtings in several US states, and compared the outcomes from our samples with the actual districts to see if there were large differences between the outcomes that would be expected from our samples and outcomes that result from the actual districts. "Geometry of Weather": Tropical cyclones are complex physical systems which exhibit coherent geometric structure. This summer, as part of a Data+ research project at Duke University, we explored ways to apply topological data analysis (TDA) ideas to complex, time varying data derived from severe weather systems. In this talk, we consider storm data from numerical simulations and explore ways of using TDA to transform descriptions of point clouds into stable, meaningful, statistical features. We will describe various aspects and difficulties of this on-going project, including data acquisition and cleaning, approximating ground-truth, and understanding existing eye detection algorithms used to identify a reference point for storm feature extraction.