MetAL: Active Semi-Supervised Learning on Graphs via Meta-Learning

Kaushalya Madhawa, Tsuyoshi Murata
Proceedings of The 12th Asian Conference on Machine Learning, PMLR 129:561-576, 2020.

Abstract

The objective of active learning (AL) is to train classification models with less labeled instances by selecting only the most informative instances for labeling. The AL algorithms designed for other data types such as images and text do not perform well on graph-structured data. Although a few heuristics-based AL algorithms have been proposed for graphs, a principled approach is lacking. In this paper, we propose MetAL, an AL approach that selects unlabeled instances that directly improve the future performance of a classification model. For a semi-supervised learning problem, we formulate the AL task as a bilevel optimization problem. Based on recent work in meta-learning, we use the meta-gradients to approximate the impact of retraining the model with any unlabeled instance on the model performance. Using multiple graph datasets belonging to different domains, we demonstrate that MetAL efficiently outperforms existing state-of-the-art AL algorithms.

Cite this Paper


BibTeX
@InProceedings{pmlr-v129-madhawa20a, title = {{M}et{A}{L}: {A}ctive {S}emi-{S}upervised {L}earning on {G}raphs via {M}eta-{L}earning}, author = {Madhawa, Kaushalya and Murata, Tsuyoshi}, booktitle = {Proceedings of The 12th Asian Conference on Machine Learning}, pages = {561--576}, year = {2020}, editor = {Pan, Sinno Jialin and Sugiyama, Masashi}, volume = {129}, series = {Proceedings of Machine Learning Research}, month = {18--20 Nov}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v129/madhawa20a/madhawa20a.pdf}, url = {https://proceedings.mlr.press/v129/madhawa20a.html}, abstract = {The objective of active learning (AL) is to train classification models with less labeled instances by selecting only the most informative instances for labeling. The AL algorithms designed for other data types such as images and text do not perform well on graph-structured data. Although a few heuristics-based AL algorithms have been proposed for graphs, a principled approach is lacking. In this paper, we propose MetAL, an AL approach that selects unlabeled instances that directly improve the future performance of a classification model. For a semi-supervised learning problem, we formulate the AL task as a bilevel optimization problem. Based on recent work in meta-learning, we use the meta-gradients to approximate the impact of retraining the model with any unlabeled instance on the model performance. Using multiple graph datasets belonging to different domains, we demonstrate that MetAL efficiently outperforms existing state-of-the-art AL algorithms.} }
Endnote
%0 Conference Paper %T MetAL: Active Semi-Supervised Learning on Graphs via Meta-Learning %A Kaushalya Madhawa %A Tsuyoshi Murata %B Proceedings of The 12th Asian Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2020 %E Sinno Jialin Pan %E Masashi Sugiyama %F pmlr-v129-madhawa20a %I PMLR %P 561--576 %U https://proceedings.mlr.press/v129/madhawa20a.html %V 129 %X The objective of active learning (AL) is to train classification models with less labeled instances by selecting only the most informative instances for labeling. The AL algorithms designed for other data types such as images and text do not perform well on graph-structured data. Although a few heuristics-based AL algorithms have been proposed for graphs, a principled approach is lacking. In this paper, we propose MetAL, an AL approach that selects unlabeled instances that directly improve the future performance of a classification model. For a semi-supervised learning problem, we formulate the AL task as a bilevel optimization problem. Based on recent work in meta-learning, we use the meta-gradients to approximate the impact of retraining the model with any unlabeled instance on the model performance. Using multiple graph datasets belonging to different domains, we demonstrate that MetAL efficiently outperforms existing state-of-the-art AL algorithms.
APA
Madhawa, K. & Murata, T.. (2020). MetAL: Active Semi-Supervised Learning on Graphs via Meta-Learning. Proceedings of The 12th Asian Conference on Machine Learning, in Proceedings of Machine Learning Research 129:561-576 Available from https://proceedings.mlr.press/v129/madhawa20a.html.

Related Material