Robust and Conjugate Gaussian Process Regression

Authors: Matias Altamirano, Francois-Xavier Briol, Jeremias Knoblauch

ICML 2024 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental To demonstrate its strong empirical performance, we deploy RCGP for problems ranging from Bayesian optimisation to sparse variational Gaussian processes.
Researcher Affiliation Academia 1Department of Statistical Science, University College London, London, United Kingdom. Correspondence to: Matias Altamirano <matias.altamirano.22@ucl.ac.uk>, Franc ois-Xavier Briol <f.briol@ucl.ac.uk>, Jeremias Knoblauch <j.knoblauch@ucl.ac.uk>.
Pseudocode No The paper contains mathematical derivations and descriptions of the method, but no explicit section or figure labeled 'Pseudocode' or 'Algorithm', nor any structured, code-like algorithm blocks.
Open Source Code Yes The code is available at https://github.com/ maltamiranomontero/RCGP.
Open Datasets Yes Boston The dataset can be found at https://www.cs.toronto.edu/ delve/data/boston/bostonDetail.html.
Dataset Splits Yes Instead of pw(y|θ, σ2), we thus maximise the leave-one-out cross validation (LOO-CV) predictive posteriors via
Hardware Specification Yes All the experiments were running on an Apple M2 Pro CPU with 16 GB of memory.
Software Dependencies No The paper mentions software like 'GPflow' and refers to 'the official implementation provided in the paper' for m-GP, but does not provide specific version numbers for these software components or their dependencies.
Experiment Setup Yes All hyperparameters are selected via L-BFGS.