You can not select more than 25 topics Topics must start with a letter or number, can include dashes ('-') and can be up to 35 characters long.
Markus Kaiser 946a5681d7 Use a version 2.8 bbl file 1 year ago
figures Remove non-essentials 1 year ago
preamble Remove ICML notices from paper 1 year ago
.gitignore Add tikz definitions; Add graphical model 1 year ago
.latexmkrc Start variational formulation 1 year ago Add readme 1 year ago
additional.bib Add Thomas' feedback 1 year ago
bayesian_warped_dependent_gp.bbl Use a version 2.8 bbl file 1 year ago
bayesian_warped_dependent_gp.tex Add cautionary note 1 year ago
icml2018.bst Start moving to icml2018 style 1 year ago
icml2018.sty Remove ICML notices from paper 1 year ago
zotero_export.bib Add Thomas' feedback 1 year ago

This paper summarizes a way of formulating priors for multiple functions using GPs. The covariance structure between these processes is assumed to be governed by convolutions over latent processes with non-linear alignment of the different outputs. Both the alignment functions and a warping after the convolutions are learned in the context of a deep GP.

Building the Document

The figures in this paper depend on a newish version of the tikz package. Everything starting with TeX Live 2016 should be fine. Additionally, since they use the graphdrawing library, the document must be built using LuaTeX. The easiest way to build the document is by running latexmk in the main directory - a .latexmkrc-file is provided with the correct configuration. If you want to use a different way of building the document, please make sure to use lualatex.