Andrew Cooper

Graduate Student, Virginia Tech
“Robust Wrapped Gaussian Process Inference for Noisy Angular Data”

Session Materials Website: Link

Speaker Bio: 

Andrew Cooper is a Ph.D. candidate in Virginia Tech’s Department of Statistics. He received his bachelors and masters degrees in Statistical Science from Duke University. His research areas include computer experiments and surrogate modeling, as well as Bayesian methodology.

Abstract: 

Angular data are commonly encountered in settings with a directional or orientational component. Regressing an angular response on real-valued features requires either intrinsically capturing the circular manifold the data lie on, or using an appropriate extrinsic transformation. A popular example of the latter is the technique of distributional wrapping, in which functions are “wrapped” around the unit circle via a modulo-2 transformation. This approach enables flexible, non-linear models like Gaussian processes (GPs) to properly account for circular structure. While straightforward in theory, the need to infer the latent unwrapped distribution along with its wrapping behavior makes inference difficult in noisy response settings, as misspecification of one can severely hinder estimation of the other. We propose a novel Bayesian approach to wrapped GP (WGP) inference that more robustly estimates the latent unwrapped space in the presence of noise compared to existing implementations. Our work is motivated by the problem of localizing radio frequency identification (RFID) tags used for tracking nuclear materials. We showcase our model’s ability to capture the relationship between frequency and phase angle in order to accurately range assets in laboratory environments.