We gratefully acknowledge support from
the Simons Foundation and member institutions.
Full-text links:

Download:

Current browse context:

stat.ML

Change to browse by:

References & Citations

Bookmark

(what is this?)
CiteULike logo BibSonomy logo Mendeley logo del.icio.us logo Digg logo Reddit logo

Statistics > Machine Learning

Title: Approximate Inference for Fully Bayesian Gaussian Process Regression

Abstract: Learning in Gaussian Process models occurs through the adaptation of hyperparameters of the mean and the covariance function. The classical approach entails maximizing the marginal likelihood yielding fixed point estimates (an approach called \textit{Type II maximum likelihood} or ML-II). An alternative learning procedure is to infer the posterior over hyperparameters in a hierarchical specification of GPs we call \textit{Fully Bayesian Gaussian Process Regression} (GPR). This work considers two approximation schemes for the intractable hyperparameter posterior: 1) Hamiltonian Monte Carlo (HMC) yielding a sampling-based approximation and 2) Variational Inference (VI) where the posterior over hyperparameters is approximated by a factorized Gaussian (mean-field) or a full-rank Gaussian accounting for correlations between hyperparameters. We analyze the predictive performance for fully Bayesian GPR on a range of benchmark data sets.
Comments: Presented at 2nd Symposium on Advances in Approximate Bayesian Inference 2019
Subjects: Machine Learning (stat.ML); Machine Learning (cs.LG)
Journal reference: Proceedings of Machine Learning Research, Volume 118 (2019) 1-12
Cite as: arXiv:1912.13440 [stat.ML]
  (or arXiv:1912.13440v2 [stat.ML] for this version)

Submission history

From: Vidhi Lalchand Miss [view email]
[v1] Tue, 31 Dec 2019 17:18:48 GMT (3087kb,D)
[v2] Mon, 6 Apr 2020 14:22:14 GMT (1542kb,D)

Link back to: arXiv, form interface, contact.