ADIOS: Architectures Deep In Output Space

Moustapha Cisse, Maruan Al-Shedivat, Samy Bengio
Proceedings of The 33rd International Conference on Machine Learning, PMLR 48:2770-2779, 2016.

Abstract

Multi-label classification is a generalization of binary classification where the task consists in predicting \emphsets of labels. With the availability of ever larger datasets, the multi-label setting has become a natural one in many applications, and the interest in solving multi-label problems has grown significantly. As expected, deep learning approaches are now yielding state-of-the-art performance for this class of problems. Unfortunately, they usually do not take into account the often unknown but nevertheless rich relationships between labels. In this paper, we propose to make use of this underlying structure by learning to partition the labels into a Markov Blanket Chain and then applying a novel deep architecture that exploits the partition. Experiments on several popular and large multi-label datasets demonstrate that our approach not only yields significant improvements, but also helps to overcome trade-offs specific to the multi-label classification setting.

Cite this Paper

BibTeX
@InProceedings{pmlr-v48-cisse16, title = {ADIOS: Architectures Deep In Output Space}, author = {Cisse, Moustapha and Al-Shedivat, Maruan and Bengio, Samy}, booktitle = {Proceedings of The 33rd International Conference on Machine Learning}, pages = {2770--2779}, year = {2016}, editor = {Balcan, Maria Florina and Weinberger, Kilian Q.}, volume = {48}, series = {Proceedings of Machine Learning Research}, address = {New York, New York, USA}, month = {20--22 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v48/cisse16.pdf}, url = {https://proceedings.mlr.press/v48/cisse16.html}, abstract = {Multi-label classification is a generalization of binary classification where the task consists in predicting \emphsets of labels. With the availability of ever larger datasets, the multi-label setting has become a natural one in many applications, and the interest in solving multi-label problems has grown significantly. As expected, deep learning approaches are now yielding state-of-the-art performance for this class of problems. Unfortunately, they usually do not take into account the often unknown but nevertheless rich relationships between labels. In this paper, we propose to make use of this underlying structure by learning to partition the labels into a Markov Blanket Chain and then applying a novel deep architecture that exploits the partition. Experiments on several popular and large multi-label datasets demonstrate that our approach not only yields significant improvements, but also helps to overcome trade-offs specific to the multi-label classification setting.} }
Endnote
%0 Conference Paper %T ADIOS: Architectures Deep In Output Space %A Moustapha Cisse %A Maruan Al-Shedivat %A Samy Bengio %B Proceedings of The 33rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2016 %E Maria Florina Balcan %E Kilian Q. Weinberger %F pmlr-v48-cisse16 %I PMLR %P 2770--2779 %U https://proceedings.mlr.press/v48/cisse16.html %V 48 %X Multi-label classification is a generalization of binary classification where the task consists in predicting \emphsets of labels. With the availability of ever larger datasets, the multi-label setting has become a natural one in many applications, and the interest in solving multi-label problems has grown significantly. As expected, deep learning approaches are now yielding state-of-the-art performance for this class of problems. Unfortunately, they usually do not take into account the often unknown but nevertheless rich relationships between labels. In this paper, we propose to make use of this underlying structure by learning to partition the labels into a Markov Blanket Chain and then applying a novel deep architecture that exploits the partition. Experiments on several popular and large multi-label datasets demonstrate that our approach not only yields significant improvements, but also helps to overcome trade-offs specific to the multi-label classification setting.
RIS
TY - CPAPER TI - ADIOS: Architectures Deep In Output Space AU - Moustapha Cisse AU - Maruan Al-Shedivat AU - Samy Bengio BT - Proceedings of The 33rd International Conference on Machine Learning DA - 2016/06/11 ED - Maria Florina Balcan ED - Kilian Q. Weinberger ID - pmlr-v48-cisse16 PB - PMLR DP - Proceedings of Machine Learning Research VL - 48 SP - 2770 EP - 2779 L1 - http://proceedings.mlr.press/v48/cisse16.pdf UR - https://proceedings.mlr.press/v48/cisse16.html AB - Multi-label classification is a generalization of binary classification where the task consists in predicting \emphsets of labels. With the availability of ever larger datasets, the multi-label setting has become a natural one in many applications, and the interest in solving multi-label problems has grown significantly. As expected, deep learning approaches are now yielding state-of-the-art performance for this class of problems. Unfortunately, they usually do not take into account the often unknown but nevertheless rich relationships between labels. In this paper, we propose to make use of this underlying structure by learning to partition the labels into a Markov Blanket Chain and then applying a novel deep architecture that exploits the partition. Experiments on several popular and large multi-label datasets demonstrate that our approach not only yields significant improvements, but also helps to overcome trade-offs specific to the multi-label classification setting. ER -
APA
Cisse, M., Al-Shedivat, M. & Bengio, S.. (2016). ADIOS: Architectures Deep In Output Space. Proceedings of The 33rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 48:2770-2779 Available from https://proceedings.mlr.press/v48/cisse16.html.

Related Material