Beau Coker

Hello there! I’m a postdoc in the Center for Theoretical Neuroscience at Columbia University’s Zuckerman Institute, working with John Cunningham on properties of variational inference. Previously, I completed a PhD in biostatistics at Harvard, advised by Finale Doshi-Velez and Brent Coull, and an MS in statistics at Duke, advised by Cynthia Rudin.

I’m interested in probabilistic machine learning, particularly Bayesian neural networks (BNNs) and Gaussian processes (GPs). I think about questions like:

• How do we design priors that encode meaningful functional properties?
• What are the theoretical connections between BNNs and GPs, especially under approximate inference?
• Recently: How can we leverage implicit regularization in probabilistic modeling?

Here’s a talk I gave at AISTATS about mean-field variational BNNs.

Contact: bc3107 [AT] columbia [DOT] edu