Zoom: https://princeton.zoom.us/j/96036181516?pwd=eVF6cGN4NXl1QXcxWm9VTWF1eTNTUT09
Meeting ID: 96036181516 Password: 510370
Efficient Gaussian process regression via function-space representations
Abstract: Over the last couple of decades a large number of numerical methods have been introduced for efficiently performing Gaussian process regression. Most of these methods focus on fast inversion of the covariance matrix that appears in the Gaussian density. In this talk I describe a slightly different approach to Gaussian process regression that relies on efficient function-space representations of Gaussian processes. These representations — fixed basis functions with Gaussian coefficients — have several substantial advantages in Gaussian process regression tasks including computational and model-interpretability benefits.
Philip Greengard is a postdoc in the statistics department at Columbia University. His research is primarily focused on developing numerical tools for efficient statistical computing. He is exploring relationships between algorithms generally associated with numerical solutions to partial differential equations and statistical environments including hierarchical modeling and Gaussian process regression. He received a PhD in Applied Mathematics from Yale where his research focused on analysis-based algorithms for scientific computing.