Reproduce “Hierarchical Inducing Point Gaussian Process for Inter-domain Observations” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Luhuan Wu, Andrew Miller, Lauren Anderson, Geoff Pleiss, David Blei, John Cunningham
HIP-GP enables inter-domain Gaussian process inference by separating transformed observations from an evenly spaced grid and exploiting its structure for cheaper covariance calculations. On 20,000 held-out observations from a simulated Milky Way-like galaxy, it achieves lower mean absolute and squared errors and higher test log likelihood than SVGP, using 62,500 versus 16,384 inducing points; this evaluates integrated observations, not dust-map accuracy.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.