Learning Structured Models with the AUC Loss and Its Generalizations

Nir Rosenfeld, Ofer Meshi, Danny Tarlow, Amir Globerson
Proceedings of the Seventeenth International Conference on Artificial Intelligence and Statistics, PMLR 33:841-849, 2014.

Abstract

Many problems involve the prediction of multiple, possibly dependent labels. The structured output prediction framework builds predictors that take these dependencies into account and use them to improve accuracy. In many such tasks, performance is evaluated by the Area Under the ROC Curve (AUC). While a framework for optimizing the AUC loss for unstructured models exists, it does not naturally extend to structured models. In this work, we propose a representation and learning formulation for optimizing structured models over the AUC loss, show how our approach generalizes the unstructured case, and provide algorithms for solving the resulting inference and learning problems. We also explore several new variants of the AUC measure which naturally arise from our formulation. Finally, we empirically show the utility of our approach in several domains.

Cite this Paper

BibTeX
@InProceedings{pmlr-v33-rosenfeld14, title = {{Learning Structured Models with the AUC Loss and Its Generalizations}}, author = {Rosenfeld, Nir and Meshi, Ofer and Tarlow, Danny and Globerson, Amir}, booktitle = {Proceedings of the Seventeenth International Conference on Artificial Intelligence and Statistics}, pages = {841--849}, year = {2014}, editor = {Kaski, Samuel and Corander, Jukka}, volume = {33}, series = {Proceedings of Machine Learning Research}, address = {Reykjavik, Iceland}, month = {22--25 Apr}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v33/rosenfeld14.pdf}, url = {https://proceedings.mlr.press/v33/rosenfeld14.html}, abstract = {Many problems involve the prediction of multiple, possibly dependent labels. The structured output prediction framework builds predictors that take these dependencies into account and use them to improve accuracy. In many such tasks, performance is evaluated by the Area Under the ROC Curve (AUC). While a framework for optimizing the AUC loss for unstructured models exists, it does not naturally extend to structured models. In this work, we propose a representation and learning formulation for optimizing structured models over the AUC loss, show how our approach generalizes the unstructured case, and provide algorithms for solving the resulting inference and learning problems. We also explore several new variants of the AUC measure which naturally arise from our formulation. Finally, we empirically show the utility of our approach in several domains.} }
Endnote
%0 Conference Paper %T Learning Structured Models with the AUC Loss and Its Generalizations %A Nir Rosenfeld %A Ofer Meshi %A Danny Tarlow %A Amir Globerson %B Proceedings of the Seventeenth International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2014 %E Samuel Kaski %E Jukka Corander %F pmlr-v33-rosenfeld14 %I PMLR %P 841--849 %U https://proceedings.mlr.press/v33/rosenfeld14.html %V 33 %X Many problems involve the prediction of multiple, possibly dependent labels. The structured output prediction framework builds predictors that take these dependencies into account and use them to improve accuracy. In many such tasks, performance is evaluated by the Area Under the ROC Curve (AUC). While a framework for optimizing the AUC loss for unstructured models exists, it does not naturally extend to structured models. In this work, we propose a representation and learning formulation for optimizing structured models over the AUC loss, show how our approach generalizes the unstructured case, and provide algorithms for solving the resulting inference and learning problems. We also explore several new variants of the AUC measure which naturally arise from our formulation. Finally, we empirically show the utility of our approach in several domains.
RIS
TY - CPAPER TI - Learning Structured Models with the AUC Loss and Its Generalizations AU - Nir Rosenfeld AU - Ofer Meshi AU - Danny Tarlow AU - Amir Globerson BT - Proceedings of the Seventeenth International Conference on Artificial Intelligence and Statistics DA - 2014/04/02 ED - Samuel Kaski ED - Jukka Corander ID - pmlr-v33-rosenfeld14 PB - PMLR DP - Proceedings of Machine Learning Research VL - 33 SP - 841 EP - 849 L1 - http://proceedings.mlr.press/v33/rosenfeld14.pdf UR - https://proceedings.mlr.press/v33/rosenfeld14.html AB - Many problems involve the prediction of multiple, possibly dependent labels. The structured output prediction framework builds predictors that take these dependencies into account and use them to improve accuracy. In many such tasks, performance is evaluated by the Area Under the ROC Curve (AUC). While a framework for optimizing the AUC loss for unstructured models exists, it does not naturally extend to structured models. In this work, we propose a representation and learning formulation for optimizing structured models over the AUC loss, show how our approach generalizes the unstructured case, and provide algorithms for solving the resulting inference and learning problems. We also explore several new variants of the AUC measure which naturally arise from our formulation. Finally, we empirically show the utility of our approach in several domains. ER -
APA
Rosenfeld, N., Meshi, O., Tarlow, D. & Globerson, A.. (2014). Learning Structured Models with the AUC Loss and Its Generalizations. Proceedings of the Seventeenth International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 33:841-849 Available from https://proceedings.mlr.press/v33/rosenfeld14.html.

Related Material