Speaker: Moon Duchin from the MGGG Redistricting Lab at Tufts. She will be speaking in-person on:
Sampling partitions, with applications to redistricting
Abstract: In the world of gerrymandering, there are many reasons to want to take a “representative sample” of the space of balanced graph partitions — that is, a sample pulled from a known distribution which is clearly relevant to the political districting problem. There are numerous Markov chain methods, plus sequential Monte Carlo, plus a direct tree decomposition method, at least. I’ll discuss this modeling problem at a high level, and will also offer some tools and diagnostics for comparisons.