Apr 21, 2025

A deep dive into Iterative Gaussian Processes

Exploring how iterative methods can make Gaussian processes scalable while properly quantifying computational uncertainty. This deep dive traces the mathematical foundations of GPs from univariate Gaussians to infinite-dimensional distributions, then introduces IterGP as a principled Bayesian approach that treats computation itself as a source of uncertainty when approximating representer weights.

A deep dive into Iterative Gaussian Processes