Show simple item record

dc.contributor.authorMa, C
dc.contributor.authorLi, Y
dc.contributor.authorHernández-Lobato, JM
dc.date.accessioned2019-07-31T11:31:49Z
dc.date.available2019-07-31T11:31:49Z
dc.date.issued2019
dc.identifier.isbn9781510886988
dc.identifier.issn2640-3498
dc.identifier.urihttps://www.repository.cam.ac.uk/handle/1810/295114
dc.description.abstractWe introduce the implicit processes (IPs), a stochastic process that places implicitly defined multivariate distributions over any finite collections of random variables. IPs are therefore highly flexible implicit priors over functions, with examples including data simulators, Bayesian neural networks and non-linear transformations of stochastic processes. A novel and efficient approximate inference algorithm for IPs, namely the variational implicit processes (VIPs), is derived using generalised wake-sleep updates. This method returns simple update equations and allows scalable hyper-parameter learning with stochastic optimization. Experiments show that VIPs return better uncertainty estimates and lower errors over existing inference methods for challenging models such as Bayesian neural networks, and Gaussian processes.
dc.language.isoen
dc.titleVariational implicit processes
dc.typeConference Object
prism.endingPage7482
prism.publicationDate2019
prism.publicationName36th International Conference on Machine Learning, ICML 2019
prism.startingPage7464
prism.volume2019-June
dc.identifier.doi10.17863/CAM.42186
rioxxterms.versionofrecord10.17863/CAM.42186
rioxxterms.versionAM
rioxxterms.licenseref.urihttp://www.rioxx.net/licenses/all-rights-reserved
rioxxterms.licenseref.startdate2019-01-01
rioxxterms.typeConference Paper/Proceeding/Abstract


Files in this item

Thumbnail

This item appears in the following Collection(s)

Show simple item record