Learning parts-based representations with nonnegative restricted boltzmann machine

© 2013 T.D. Nguyen, T. Tran, D. Phung & S. Venkatesh.The success of any machine learning system depends critically on effective representations of data. In many cases, especially those in vision, it is desirable that a representation scheme uncovers the parts-based, additive nature of the data....

Full description

Bibliographic Details
Main Authors: Nguyeny, T., Tran, The Truyen, Phungy, D., Venkateshy, S.
Format: Conference Paper
Published: 2013
Online Access:http://hdl.handle.net/20.500.11937/51761
_version_ 1848758763804164096
author Nguyeny, T.
Tran, The Truyen
Phungy, D.
Venkateshy, S.
author_facet Nguyeny, T.
Tran, The Truyen
Phungy, D.
Venkateshy, S.
author_sort Nguyeny, T.
building Curtin Institutional Repository
collection Online Access
description © 2013 T.D. Nguyen, T. Tran, D. Phung & S. Venkatesh.The success of any machine learning system depends critically on effective representations of data. In many cases, especially those in vision, it is desirable that a representation scheme uncovers the parts-based, additive nature of the data. Of current representation learning schemes, restricted Boltzmann machines (RBMs) have proved to be highly effective in unsupervised settings. However, when it comes to parts-based discovery, RBMs do not usually produce satisfactory results. We enhance such capacity of RBMs by introducing nonnegativity into the model weights, resulting in a variant called nonnegative restricted Boltzmann machine (NRBM). The NRBM produces not only controllable decomposition of data into interpretable parts but also offers a way to estimate the intrinsic nonlinear dimensionality of data. We demonstrate the capacity of our model on well-known datasets of handwritten digits, faces and documents. The decomposition quality on images is comparable with or better than what produced by the nonnegative matrix factorisation (NMF), and the thematic features uncovered from text are qualitatively interpretable in a similar manner to that of the latent Dirichlet allocation (LDA). However, the learnt features, when used for classification, are more discriminative than those discovered by both NMF and LDA and comparable with those by RBM.
first_indexed 2025-11-14T09:49:10Z
format Conference Paper
id curtin-20.500.11937-51761
institution Curtin University Malaysia
institution_category Local University
last_indexed 2025-11-14T09:49:10Z
publishDate 2013
recordtype eprints
repository_type Digital Repository
spelling curtin-20.500.11937-517612017-04-04T02:46:23Z Learning parts-based representations with nonnegative restricted boltzmann machine Nguyeny, T. Tran, The Truyen Phungy, D. Venkateshy, S. © 2013 T.D. Nguyen, T. Tran, D. Phung & S. Venkatesh.The success of any machine learning system depends critically on effective representations of data. In many cases, especially those in vision, it is desirable that a representation scheme uncovers the parts-based, additive nature of the data. Of current representation learning schemes, restricted Boltzmann machines (RBMs) have proved to be highly effective in unsupervised settings. However, when it comes to parts-based discovery, RBMs do not usually produce satisfactory results. We enhance such capacity of RBMs by introducing nonnegativity into the model weights, resulting in a variant called nonnegative restricted Boltzmann machine (NRBM). The NRBM produces not only controllable decomposition of data into interpretable parts but also offers a way to estimate the intrinsic nonlinear dimensionality of data. We demonstrate the capacity of our model on well-known datasets of handwritten digits, faces and documents. The decomposition quality on images is comparable with or better than what produced by the nonnegative matrix factorisation (NMF), and the thematic features uncovered from text are qualitatively interpretable in a similar manner to that of the latent Dirichlet allocation (LDA). However, the learnt features, when used for classification, are more discriminative than those discovered by both NMF and LDA and comparable with those by RBM. 2013 Conference Paper http://hdl.handle.net/20.500.11937/51761 restricted
spellingShingle Nguyeny, T.
Tran, The Truyen
Phungy, D.
Venkateshy, S.
Learning parts-based representations with nonnegative restricted boltzmann machine
title Learning parts-based representations with nonnegative restricted boltzmann machine
title_full Learning parts-based representations with nonnegative restricted boltzmann machine
title_fullStr Learning parts-based representations with nonnegative restricted boltzmann machine
title_full_unstemmed Learning parts-based representations with nonnegative restricted boltzmann machine
title_short Learning parts-based representations with nonnegative restricted boltzmann machine
title_sort learning parts-based representations with nonnegative restricted boltzmann machine
url http://hdl.handle.net/20.500.11937/51761