About the event

Department of Mathematics and Statistics

Seminar in Theory and Applications of Discrete Math, Linear Algebra and Number Theory

Case Studies of Short Burst Optimization for Multiple Redistricting Criteria

Phousawanh Peaungvongpakdy

Abstract: In complex and non-convex energy landscapes like those that occur in the discrete context of partition sampling for redistricting, optimal states can be difficult to discover. In 2023 Cannon et al. demonstrated that “short bursts’’, biased random walks performed in a small number of steps and repeated, led to significantly better performance than simple biased walks in finding global optima in this redistricting context. In this talk, we will present case studies applying short bursts to a variety of traditional redistricting criteria. Applying this method on real-world redistricting data allows us to explore tradeoffs between metrics and the corresponding Pareto frontiers. We also present some results about short bursts on simple random walks on labeled graphs.