Learning Ordered Representations with Nested Dropout

Oren Rippel, Michael Gelbart, Ryan Adams
Proceedings of the 31st International Conference on Machine Learning, PMLR 32(2):1746-1754, 2014.

Abstract

In this paper, we present results on ordered representations of data in which different dimensions have different degrees of importance. To learn these representations we introduce nested dropout, a procedure for stochastically removing coherent nested sets of hidden units in a neural network. We first present a sequence of theoretical results in the simple case of a semi-linear autoencoder. We rigorously show that the application of nested dropout enforces identifiability of the units, which leads to an exact equivalence with PCA. We then extend the algorithm to deep models and demonstrate the relevance of ordered representations to a number of applications. Specifically, we use the ordered property of the learned codes to construct hash-based data structures that permit very fast retrieval, achieving retrieval in time logarithmic in the database size and independent of the dimensionality of the representation. This allows the use of codes that are hundreds of times longer than currently feasible for retrieval. We therefore avoid the diminished quality associated with short codes, while still performing retrieval that is competitive in speed with existing methods. We also show that ordered representations are a promising way to learn adaptive compression for efficient online data reconstruction.

Cite this Paper


BibTeX
@InProceedings{pmlr-v32-rippel14, title = {Learning Ordered Representations with Nested Dropout}, author = {Rippel, Oren and Gelbart, Michael and Adams, Ryan}, booktitle = {Proceedings of the 31st International Conference on Machine Learning}, pages = {1746--1754}, year = {2014}, editor = {Xing, Eric P. and Jebara, Tony}, volume = {32}, number = {2}, series = {Proceedings of Machine Learning Research}, address = {Bejing, China}, month = {22--24 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v32/rippel14.pdf}, url = {https://proceedings.mlr.press/v32/rippel14.html}, abstract = {In this paper, we present results on ordered representations of data in which different dimensions have different degrees of importance. To learn these representations we introduce nested dropout, a procedure for stochastically removing coherent nested sets of hidden units in a neural network. We first present a sequence of theoretical results in the simple case of a semi-linear autoencoder. We rigorously show that the application of nested dropout enforces identifiability of the units, which leads to an exact equivalence with PCA. We then extend the algorithm to deep models and demonstrate the relevance of ordered representations to a number of applications. Specifically, we use the ordered property of the learned codes to construct hash-based data structures that permit very fast retrieval, achieving retrieval in time logarithmic in the database size and independent of the dimensionality of the representation. This allows the use of codes that are hundreds of times longer than currently feasible for retrieval. We therefore avoid the diminished quality associated with short codes, while still performing retrieval that is competitive in speed with existing methods. We also show that ordered representations are a promising way to learn adaptive compression for efficient online data reconstruction.} }
Endnote
%0 Conference Paper %T Learning Ordered Representations with Nested Dropout %A Oren Rippel %A Michael Gelbart %A Ryan Adams %B Proceedings of the 31st International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2014 %E Eric P. Xing %E Tony Jebara %F pmlr-v32-rippel14 %I PMLR %P 1746--1754 %U https://proceedings.mlr.press/v32/rippel14.html %V 32 %N 2 %X In this paper, we present results on ordered representations of data in which different dimensions have different degrees of importance. To learn these representations we introduce nested dropout, a procedure for stochastically removing coherent nested sets of hidden units in a neural network. We first present a sequence of theoretical results in the simple case of a semi-linear autoencoder. We rigorously show that the application of nested dropout enforces identifiability of the units, which leads to an exact equivalence with PCA. We then extend the algorithm to deep models and demonstrate the relevance of ordered representations to a number of applications. Specifically, we use the ordered property of the learned codes to construct hash-based data structures that permit very fast retrieval, achieving retrieval in time logarithmic in the database size and independent of the dimensionality of the representation. This allows the use of codes that are hundreds of times longer than currently feasible for retrieval. We therefore avoid the diminished quality associated with short codes, while still performing retrieval that is competitive in speed with existing methods. We also show that ordered representations are a promising way to learn adaptive compression for efficient online data reconstruction.
RIS
TY - CPAPER TI - Learning Ordered Representations with Nested Dropout AU - Oren Rippel AU - Michael Gelbart AU - Ryan Adams BT - Proceedings of the 31st International Conference on Machine Learning DA - 2014/06/18 ED - Eric P. Xing ED - Tony Jebara ID - pmlr-v32-rippel14 PB - PMLR DP - Proceedings of Machine Learning Research VL - 32 IS - 2 SP - 1746 EP - 1754 L1 - http://proceedings.mlr.press/v32/rippel14.pdf UR - https://proceedings.mlr.press/v32/rippel14.html AB - In this paper, we present results on ordered representations of data in which different dimensions have different degrees of importance. To learn these representations we introduce nested dropout, a procedure for stochastically removing coherent nested sets of hidden units in a neural network. We first present a sequence of theoretical results in the simple case of a semi-linear autoencoder. We rigorously show that the application of nested dropout enforces identifiability of the units, which leads to an exact equivalence with PCA. We then extend the algorithm to deep models and demonstrate the relevance of ordered representations to a number of applications. Specifically, we use the ordered property of the learned codes to construct hash-based data structures that permit very fast retrieval, achieving retrieval in time logarithmic in the database size and independent of the dimensionality of the representation. This allows the use of codes that are hundreds of times longer than currently feasible for retrieval. We therefore avoid the diminished quality associated with short codes, while still performing retrieval that is competitive in speed with existing methods. We also show that ordered representations are a promising way to learn adaptive compression for efficient online data reconstruction. ER -
APA
Rippel, O., Gelbart, M. & Adams, R.. (2014). Learning Ordered Representations with Nested Dropout. Proceedings of the 31st International Conference on Machine Learning, in Proceedings of Machine Learning Research 32(2):1746-1754 Available from https://proceedings.mlr.press/v32/rippel14.html.

Related Material