Damianou, A.C. and Lawrence, N.D. orcid.org/0000-0001-9258-1030 (2013) Deep Gaussian Processes. In: Proceedings of the Sixteenth International Conference on Artificial Intelligence and Statistics. Sixteenth International Conference on Artificial Intelligence and Statistics, 29 Apr - 01 May 2013, Scottsdale, AZ, USA. JMLR Workshop and Conference Proceedings, 31 . JMLR
Abstract
In this paper we introduce deep Gaussian process (GP) models. Deep GPs are a deep belief network based on Gaussian process mappings. The data is modeled as the output of a multivariate GP. The inputs to that Gaussian process are then governed by another GP. A single layer model is equivalent to a standard GP or the GP latent variable model (GP-LVM). We perform inference in the model by approximate variational marginalization. This results in a strict lower bound on the marginal likelihood of the model which we use for model selection (number of layers and nodes per layer). Deep belief networks are typically applied to relatively large data sets using stochastic gradient descent for optimization. Our fully Bayesian treatment allows for the application of deep models even when data is scarce. Model selection by our variational bound shows that a five layer hierarchy is justified even when modelling a digit data set containing only 150 examples.
Metadata
Item Type: | Proceedings Paper |
---|---|
Authors/Creators: |
|
Copyright, Publisher and Additional Information: | © 2013 The Author(s) |
Keywords: | stat.ML; stat.ML; cs.LG; math.PR; 60G15, 58E30; G.3; G.1.2; I.2.6 |
Dates: |
|
Institution: | The University of Sheffield |
Academic Units: | The University of Sheffield > Faculty of Engineering (Sheffield) > Department of Computer Science (Sheffield) The University of Sheffield > Faculty of Medicine, Dentistry and Health (Sheffield) > Department of Neuroscience (Sheffield) |
Depositing User: | Symplectic Sheffield |
Date Deposited: | 05 Oct 2016 15:37 |
Last Modified: | 24 Dec 2022 02:02 |
Published Version: | http://www.jmlr.org/proceedings/papers/v31/damiano... |
Status: | Published |
Publisher: | JMLR |
Series Name: | JMLR Workshop and Conference Proceedings |
Refereed: | Yes |
Related URLs: | |
Open Archives Initiative ID (OAI ID): | oai:eprints.whiterose.ac.uk:103731 |