Learning Context-dependent Label Permutations for Multi-label Classification

Jinseok Nam, Young-Bum Kim, Eneldo Loza Mencia, Sunghyun Park, Ruhi Sarikaya, Johannes Fürnkranz
Proceedings of the 36th International Conference on Machine Learning, PMLR 97:4733-4742, 2019.

Abstract

A key problem in multi-label classification is to utilize dependencies among the labels. Chaining classifiers are a simple technique for addressing this problem but current algorithms all assume a fixed, static label ordering. In this work, we propose a multi-label classification approach which allows to choose a dynamic, context-dependent label ordering. Our proposed approach consists of two sub-components: a simple EM-like algorithm which bootstraps the learned model, and a more elaborate approach based on reinforcement learning. Our experiments on three public multi-label classification benchmarks show that our proposed dynamic label ordering approach based on reinforcement learning outperforms recurrent neural networks with fixed label ordering across both bipartition and ranking measures on all the three datasets. As a result, we obtain a powerful sequence prediction-based algorithm for multi-label classification, which is able to efficiently and explicitly exploit label dependencies.

Cite this Paper

BibTeX
@InProceedings{pmlr-v97-nam19a, title = {Learning Context-dependent Label Permutations for Multi-label Classification}, author = {Nam, Jinseok and Kim, Young-Bum and Mencia, Eneldo Loza and Park, Sunghyun and Sarikaya, Ruhi and F{\"u}rnkranz, Johannes}, booktitle = {Proceedings of the 36th International Conference on Machine Learning}, pages = {4733--4742}, year = {2019}, editor = {Chaudhuri, Kamalika and Salakhutdinov, Ruslan}, volume = {97}, series = {Proceedings of Machine Learning Research}, month = {09--15 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v97/nam19a/nam19a.pdf}, url = {https://proceedings.mlr.press/v97/nam19a.html}, abstract = {A key problem in multi-label classification is to utilize dependencies among the labels. Chaining classifiers are a simple technique for addressing this problem but current algorithms all assume a fixed, static label ordering. In this work, we propose a multi-label classification approach which allows to choose a dynamic, context-dependent label ordering. Our proposed approach consists of two sub-components: a simple EM-like algorithm which bootstraps the learned model, and a more elaborate approach based on reinforcement learning. Our experiments on three public multi-label classification benchmarks show that our proposed dynamic label ordering approach based on reinforcement learning outperforms recurrent neural networks with fixed label ordering across both bipartition and ranking measures on all the three datasets. As a result, we obtain a powerful sequence prediction-based algorithm for multi-label classification, which is able to efficiently and explicitly exploit label dependencies.} }
Endnote
%0 Conference Paper %T Learning Context-dependent Label Permutations for Multi-label Classification %A Jinseok Nam %A Young-Bum Kim %A Eneldo Loza Mencia %A Sunghyun Park %A Ruhi Sarikaya %A Johannes Fürnkranz %B Proceedings of the 36th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2019 %E Kamalika Chaudhuri %E Ruslan Salakhutdinov %F pmlr-v97-nam19a %I PMLR %P 4733--4742 %U https://proceedings.mlr.press/v97/nam19a.html %V 97 %X A key problem in multi-label classification is to utilize dependencies among the labels. Chaining classifiers are a simple technique for addressing this problem but current algorithms all assume a fixed, static label ordering. In this work, we propose a multi-label classification approach which allows to choose a dynamic, context-dependent label ordering. Our proposed approach consists of two sub-components: a simple EM-like algorithm which bootstraps the learned model, and a more elaborate approach based on reinforcement learning. Our experiments on three public multi-label classification benchmarks show that our proposed dynamic label ordering approach based on reinforcement learning outperforms recurrent neural networks with fixed label ordering across both bipartition and ranking measures on all the three datasets. As a result, we obtain a powerful sequence prediction-based algorithm for multi-label classification, which is able to efficiently and explicitly exploit label dependencies.
APA
Nam, J., Kim, Y., Mencia, E.L., Park, S., Sarikaya, R. & Fürnkranz, J.. (2019). Learning Context-dependent Label Permutations for Multi-label Classification. Proceedings of the 36th International Conference on Machine Learning, in Proceedings of Machine Learning Research 97:4733-4742 Available from https://proceedings.mlr.press/v97/nam19a.html.

Related Material