The Gaussian Process Density Sampler
Abstract
We present the Gaussian Process Density Sampler (GPDS), an exchangeable generative model for use in nonparametric Bayesian density estimation. Samples drawn from the GPDS are consistent with exact, independent samples from a fixed density function that is a transformation of a function drawn from a Gaussian process prior. Our formulation allows us to infer an unknown density from data using Markov chain Monte Carlo, which gives samples from the posterior distribution over density functions and from the predictive distribution on data space. We can also infer the hyperparameters of the Gaussian process. We compare this density modeling technique to several existing techniques on a toy problem and a skull-reconstruction task.
Cite
Text
Murray et al. "The Gaussian Process Density Sampler." Neural Information Processing Systems, 2008.Markdown
[Murray et al. "The Gaussian Process Density Sampler." Neural Information Processing Systems, 2008.](https://mlanthology.org/neurips/2008/murray2008neurips-gaussian/)BibTeX
@inproceedings{murray2008neurips-gaussian,
title = {{The Gaussian Process Density Sampler}},
author = {Murray, Iain and MacKay, David and Adams, Ryan P.},
booktitle = {Neural Information Processing Systems},
year = {2008},
pages = {9-16},
url = {https://mlanthology.org/neurips/2008/murray2008neurips-gaussian/}
}