Repository logo
 

Convolutional Gaussian Processes

Accepted version
Peer-reviewed

Type

Conference Object

Change log

Authors

van der Wilk, Mark 
Rasmussen, Carl Edward 
Hensman, James 

Abstract

We present a practical way of introducing convolutional structure into Gaussian processes, making them more suited to high-dimensional inputs like images. The main contribution of our work is the construction of an inter-domain inducing point approximation that is well-tailored to the convolutional kernel. This allows us to gain the generalisation benefit of a convolutional kernel, together with fast but accurate posterior inference. We investigate several variations of the convolutional kernel, and apply it to MNIST and CIFAR-10, which have both been known to be challenging for Gaussian processes. We also show how the marginal likelihood can be used to find an optimal weighting between convolutional and RBF kernels to further improve performance. We hope that this illustration of the usefulness of a marginal likelihood will help automate discovering architectures in larger models.

Description

Keywords

Journal Title

Proceedings of the 31st International Conference on Neural Information Processing Systems

Conference Name

Neural Information Processing Systems 2017 (NIPS'17)

Journal ISSN

1049-5258

Volume Title

Publisher

Curran Associates Inc.
Sponsorship
Engineering and Physical Sciences Research Council (EP/J012300/1)
Alan Turing Institute (unknown)