Multiresolution Deep Belief Networks

Yichuan Tang, Abdel-Rahman Mohamed
Proceedings of the Fifteenth International Conference on Artificial Intelligence and Statistics, PMLR 22:1203-1211, 2012.

Abstract

Motivated by the observation that coarse and fine resolutions of an image reveal different structures in the underlying visual phenomenon, we present a model based on the Deep Belief Network (DBN) which learns features from the multiscale representation of images. A Laplacian Pyramid is first constructed for each image. DBNs are then trained separately at each level of the pyramid. Finally, a top level RBM combines these DBNs into a single network we call the Multiresolution Deep Belief Network (MrDBN). Experiments show that MrDBNs generalize better than standard DBNs on NORB classification and TIMIT phone recognition. In the domain of generative learning, we demonstrate the superiority of MrDBNs at modeling face images.

Cite this Paper


BibTeX
@InProceedings{pmlr-v22-tang12, title = {Multiresolution Deep Belief Networks}, author = {Tang, Yichuan and Mohamed, Abdel-Rahman}, booktitle = {Proceedings of the Fifteenth International Conference on Artificial Intelligence and Statistics}, pages = {1203--1211}, year = {2012}, editor = {Lawrence, Neil D. and Girolami, Mark}, volume = {22}, series = {Proceedings of Machine Learning Research}, address = {La Palma, Canary Islands}, month = {21--23 Apr}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v22/tang12/tang12.pdf}, url = {https://proceedings.mlr.press/v22/tang12.html}, abstract = {Motivated by the observation that coarse and fine resolutions of an image reveal different structures in the underlying visual phenomenon, we present a model based on the Deep Belief Network (DBN) which learns features from the multiscale representation of images. A Laplacian Pyramid is first constructed for each image. DBNs are then trained separately at each level of the pyramid. Finally, a top level RBM combines these DBNs into a single network we call the Multiresolution Deep Belief Network (MrDBN). Experiments show that MrDBNs generalize better than standard DBNs on NORB classification and TIMIT phone recognition. In the domain of generative learning, we demonstrate the superiority of MrDBNs at modeling face images.} }
Endnote
%0 Conference Paper %T Multiresolution Deep Belief Networks %A Yichuan Tang %A Abdel-Rahman Mohamed %B Proceedings of the Fifteenth International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2012 %E Neil D. Lawrence %E Mark Girolami %F pmlr-v22-tang12 %I PMLR %P 1203--1211 %U https://proceedings.mlr.press/v22/tang12.html %V 22 %X Motivated by the observation that coarse and fine resolutions of an image reveal different structures in the underlying visual phenomenon, we present a model based on the Deep Belief Network (DBN) which learns features from the multiscale representation of images. A Laplacian Pyramid is first constructed for each image. DBNs are then trained separately at each level of the pyramid. Finally, a top level RBM combines these DBNs into a single network we call the Multiresolution Deep Belief Network (MrDBN). Experiments show that MrDBNs generalize better than standard DBNs on NORB classification and TIMIT phone recognition. In the domain of generative learning, we demonstrate the superiority of MrDBNs at modeling face images.
RIS
TY - CPAPER TI - Multiresolution Deep Belief Networks AU - Yichuan Tang AU - Abdel-Rahman Mohamed BT - Proceedings of the Fifteenth International Conference on Artificial Intelligence and Statistics DA - 2012/03/21 ED - Neil D. Lawrence ED - Mark Girolami ID - pmlr-v22-tang12 PB - PMLR DP - Proceedings of Machine Learning Research VL - 22 SP - 1203 EP - 1211 L1 - http://proceedings.mlr.press/v22/tang12/tang12.pdf UR - https://proceedings.mlr.press/v22/tang12.html AB - Motivated by the observation that coarse and fine resolutions of an image reveal different structures in the underlying visual phenomenon, we present a model based on the Deep Belief Network (DBN) which learns features from the multiscale representation of images. A Laplacian Pyramid is first constructed for each image. DBNs are then trained separately at each level of the pyramid. Finally, a top level RBM combines these DBNs into a single network we call the Multiresolution Deep Belief Network (MrDBN). Experiments show that MrDBNs generalize better than standard DBNs on NORB classification and TIMIT phone recognition. In the domain of generative learning, we demonstrate the superiority of MrDBNs at modeling face images. ER -
APA
Tang, Y. & Mohamed, A.. (2012). Multiresolution Deep Belief Networks. Proceedings of the Fifteenth International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 22:1203-1211 Available from https://proceedings.mlr.press/v22/tang12.html.

Related Material