Unsupervised discovery of nonlinear structure using contrastive backpropagation.

We describe a way of modeling high-dimensional data vectors by using an unsupervised, nonlinear, multilayer neural network in which the activity of each neuron-like unit makes an additive contribution to a global energy score that indicates how surprised the network is by the data vector. The connec...

Full beskrivning

Bibliografiska uppgifter
Huvudupphovsmän: Hinton, G, Osindero, S, Welling, M, Teh, Y
Materialtyp: Journal article
Språk:English
Publicerad: 2006
_version_ 1826281304585404416
author Hinton, G
Osindero, S
Welling, M
Teh, Y
author_facet Hinton, G
Osindero, S
Welling, M
Teh, Y
author_sort Hinton, G
collection OXFORD
description We describe a way of modeling high-dimensional data vectors by using an unsupervised, nonlinear, multilayer neural network in which the activity of each neuron-like unit makes an additive contribution to a global energy score that indicates how surprised the network is by the data vector. The connection weights that determine how the activity of each unit depends on the activities in earlier layers are learned by minimizing the energy assigned to data vectors that are actually observed and maximizing the energy assigned to "confabulations" that are generated by perturbing an observed data vector in a direction that decreases its energy under the current model.
first_indexed 2024-03-07T00:26:48Z
format Journal article
id oxford-uuid:7e696a80-79be-4bc6-a78c-834b6ef1a1ce
institution University of Oxford
language English
last_indexed 2024-03-07T00:26:48Z
publishDate 2006
record_format dspace
spelling oxford-uuid:7e696a80-79be-4bc6-a78c-834b6ef1a1ce2022-03-26T21:09:58ZUnsupervised discovery of nonlinear structure using contrastive backpropagation.Journal articlehttp://purl.org/coar/resource_type/c_dcae04bcuuid:7e696a80-79be-4bc6-a78c-834b6ef1a1ceEnglishSymplectic Elements at Oxford2006Hinton, GOsindero, SWelling, MTeh, YWe describe a way of modeling high-dimensional data vectors by using an unsupervised, nonlinear, multilayer neural network in which the activity of each neuron-like unit makes an additive contribution to a global energy score that indicates how surprised the network is by the data vector. The connection weights that determine how the activity of each unit depends on the activities in earlier layers are learned by minimizing the energy assigned to data vectors that are actually observed and maximizing the energy assigned to "confabulations" that are generated by perturbing an observed data vector in a direction that decreases its energy under the current model.
spellingShingle Hinton, G
Osindero, S
Welling, M
Teh, Y
Unsupervised discovery of nonlinear structure using contrastive backpropagation.
title Unsupervised discovery of nonlinear structure using contrastive backpropagation.
title_full Unsupervised discovery of nonlinear structure using contrastive backpropagation.
title_fullStr Unsupervised discovery of nonlinear structure using contrastive backpropagation.
title_full_unstemmed Unsupervised discovery of nonlinear structure using contrastive backpropagation.
title_short Unsupervised discovery of nonlinear structure using contrastive backpropagation.
title_sort unsupervised discovery of nonlinear structure using contrastive backpropagation
work_keys_str_mv AT hintong unsuperviseddiscoveryofnonlinearstructureusingcontrastivebackpropagation
AT osinderos unsuperviseddiscoveryofnonlinearstructureusingcontrastivebackpropagation
AT wellingm unsuperviseddiscoveryofnonlinearstructureusingcontrastivebackpropagation
AT tehy unsuperviseddiscoveryofnonlinearstructureusingcontrastivebackpropagation