Abstract: Every ten years the United States performs a census which determines how many members of congress will represent each state. Then begins an unfortunate battle, as cartographers manipulate the boundaries for political power in a process called gerrymandering. For instance, a city can become one strong Democratic hold out, or divided up creating multiple more moderate districts.
Many researchers have tried to find ways to prevent gerrymandering. They have suggested ways to detect it by looking at the boundaries of the districts for irregularity, or by looking at the results over time to detect a bias towards one party. There has also been increasing interest for ways to generate the maps automatically.
In this talk, we present a a new class of districting algorithms based on discrete weighted Voronoi regions. These algorithms feature an iterative updating of the distances in order to balance district compactness and population as much as possible.
This talk is based on two papers: