Reproduce “A Framework for Evaluating Approximation Methods for Gaussian Process Regression” 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
Krzysztof Chalupka, Christopher K. I. Williams, Iain Murray
The authors evaluate Gaussian process approximations by comparing prediction error with CPU time across computational phases. They test Subset of Data (SoD), FITC, Hybrid, and Local methods on four datasets, timing hyperparameter learning and prediction. SoD outperforms FITC at matched hyperparameter-training time, while FITC has better accuracy at matched test time; this pattern is limited to tested datasets and implementations.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.