Optimizing the F-Measure in Multi-Label Classification: Plug-in Rule Approach versus Structured Loss Minimization

Krzysztof Dembczynski, Arkadiusz Jachnik, Wojciech Kotlowski, Willem Waegeman, Eyke Huellermeier
Proceedings of the 30th International Conference on Machine Learning, PMLR 28(3):1130-1138, 2013.

Abstract

We compare the plug-in rule approach for optimizing the F-measure in multi-label classification with an approach based on structured loss minimization, such as the structured support vector machine (SSVM). Whereas the former derives an optimal prediction from a probabilistic model in a separate inference step, the latter seeks to optimize the F-measure directly during the training phase. We introduce a novel plug-in rule algorithm that estimates all parameters required for a Bayes-optimal prediction via a set of multinomial regression models, and we compare this algorithm with SSVMs in terms of computational complexity and statistical consistency. As a main theoretical result, we show that our plug-in rule algorithm is consistent, whereas the SSVM approaches are not. Finally, we present results of a large experimental study showing the benefits of the introduced algorithm.

Cite this Paper

BibTeX
@InProceedings{pmlr-v28-dembczynski13, title = {Optimizing the F-Measure in Multi-Label Classification: Plug-in Rule Approach versus Structured Loss Minimization}, author = {Dembczynski, Krzysztof and Jachnik, Arkadiusz and Kotlowski, Wojciech and Waegeman, Willem and Huellermeier, Eyke}, booktitle = {Proceedings of the 30th International Conference on Machine Learning}, pages = {1130--1138}, year = {2013}, editor = {Dasgupta, Sanjoy and McAllester, David}, volume = {28}, number = {3}, series = {Proceedings of Machine Learning Research}, address = {Atlanta, Georgia, USA}, month = {17--19 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v28/dembczynski13.pdf}, url = {https://proceedings.mlr.press/v28/dembczynski13.html}, abstract = {We compare the plug-in rule approach for optimizing the F-measure in multi-label classification with an approach based on structured loss minimization, such as the structured support vector machine (SSVM). Whereas the former derives an optimal prediction from a probabilistic model in a separate inference step, the latter seeks to optimize the F-measure directly during the training phase. We introduce a novel plug-in rule algorithm that estimates all parameters required for a Bayes-optimal prediction via a set of multinomial regression models, and we compare this algorithm with SSVMs in terms of computational complexity and statistical consistency. As a main theoretical result, we show that our plug-in rule algorithm is consistent, whereas the SSVM approaches are not. Finally, we present results of a large experimental study showing the benefits of the introduced algorithm.} }
Endnote
%0 Conference Paper %T Optimizing the F-Measure in Multi-Label Classification: Plug-in Rule Approach versus Structured Loss Minimization %A Krzysztof Dembczynski %A Arkadiusz Jachnik %A Wojciech Kotlowski %A Willem Waegeman %A Eyke Huellermeier %B Proceedings of the 30th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2013 %E Sanjoy Dasgupta %E David McAllester %F pmlr-v28-dembczynski13 %I PMLR %P 1130--1138 %U https://proceedings.mlr.press/v28/dembczynski13.html %V 28 %N 3 %X We compare the plug-in rule approach for optimizing the F-measure in multi-label classification with an approach based on structured loss minimization, such as the structured support vector machine (SSVM). Whereas the former derives an optimal prediction from a probabilistic model in a separate inference step, the latter seeks to optimize the F-measure directly during the training phase. We introduce a novel plug-in rule algorithm that estimates all parameters required for a Bayes-optimal prediction via a set of multinomial regression models, and we compare this algorithm with SSVMs in terms of computational complexity and statistical consistency. As a main theoretical result, we show that our plug-in rule algorithm is consistent, whereas the SSVM approaches are not. Finally, we present results of a large experimental study showing the benefits of the introduced algorithm.
RIS
TY - CPAPER TI - Optimizing the F-Measure in Multi-Label Classification: Plug-in Rule Approach versus Structured Loss Minimization AU - Krzysztof Dembczynski AU - Arkadiusz Jachnik AU - Wojciech Kotlowski AU - Willem Waegeman AU - Eyke Huellermeier BT - Proceedings of the 30th International Conference on Machine Learning DA - 2013/05/26 ED - Sanjoy Dasgupta ED - David McAllester ID - pmlr-v28-dembczynski13 PB - PMLR DP - Proceedings of Machine Learning Research VL - 28 IS - 3 SP - 1130 EP - 1138 L1 - http://proceedings.mlr.press/v28/dembczynski13.pdf UR - https://proceedings.mlr.press/v28/dembczynski13.html AB - We compare the plug-in rule approach for optimizing the F-measure in multi-label classification with an approach based on structured loss minimization, such as the structured support vector machine (SSVM). Whereas the former derives an optimal prediction from a probabilistic model in a separate inference step, the latter seeks to optimize the F-measure directly during the training phase. We introduce a novel plug-in rule algorithm that estimates all parameters required for a Bayes-optimal prediction via a set of multinomial regression models, and we compare this algorithm with SSVMs in terms of computational complexity and statistical consistency. As a main theoretical result, we show that our plug-in rule algorithm is consistent, whereas the SSVM approaches are not. Finally, we present results of a large experimental study showing the benefits of the introduced algorithm. ER -
APA
Dembczynski, K., Jachnik, A., Kotlowski, W., Waegeman, W. & Huellermeier, E.. (2013). Optimizing the F-Measure in Multi-Label Classification: Plug-in Rule Approach versus Structured Loss Minimization. Proceedings of the 30th International Conference on Machine Learning, in Proceedings of Machine Learning Research 28(3):1130-1138 Available from https://proceedings.mlr.press/v28/dembczynski13.html.

Related Material