Progressive Graph Learning for Open-Set Domain Adaptation

Yadan Luo, Zijian Wang, Zi Huang, Mahsa Baktashmotlagh
Proceedings of the 37th International Conference on Machine Learning, PMLR 119:6468-6478, 2020.

Abstract

Domain shift is a fundamental problem in visual recognition which typically arises when the source and target data follow different distributions. The existing domain adaptation approaches which tackle this problem work in the "closed-set" setting with the assumption that the source and the target data share exactly the same classes of objects. In this paper, we tackle a more realistic problem of the "open-set" domain shift where the target data contains additional classes that were not present in the source data. More specifically, we introduce an end-to-end Progressive Graph Learning (PGL) framework where a graph neural network with episodic training is integrated to suppress underlying conditional shift and adversarial learning is adopted to close the gap between the source and target distributions. Compared to the existing open-set adaptation approaches, our approach guarantees to achieve a tighter upper bound of the target error. Extensive experiments on three standard open-set benchmarks evidence that our approach significantly outperforms the state-of-the-arts in open-set domain adaptation.

Cite this Paper

BibTeX
@InProceedings{pmlr-v119-luo20b, title = {Progressive Graph Learning for Open-Set Domain Adaptation}, author = {Luo, Yadan and Wang, Zijian and Huang, Zi and Baktashmotlagh, Mahsa}, booktitle = {Proceedings of the 37th International Conference on Machine Learning}, pages = {6468--6478}, year = {2020}, editor = {III, Hal Daumé and Singh, Aarti}, volume = {119}, series = {Proceedings of Machine Learning Research}, month = {13--18 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v119/luo20b/luo20b.pdf}, url = {https://proceedings.mlr.press/v119/luo20b.html}, abstract = {Domain shift is a fundamental problem in visual recognition which typically arises when the source and target data follow different distributions. The existing domain adaptation approaches which tackle this problem work in the "closed-set" setting with the assumption that the source and the target data share exactly the same classes of objects. In this paper, we tackle a more realistic problem of the "open-set" domain shift where the target data contains additional classes that were not present in the source data. More specifically, we introduce an end-to-end Progressive Graph Learning (PGL) framework where a graph neural network with episodic training is integrated to suppress underlying conditional shift and adversarial learning is adopted to close the gap between the source and target distributions. Compared to the existing open-set adaptation approaches, our approach guarantees to achieve a tighter upper bound of the target error. Extensive experiments on three standard open-set benchmarks evidence that our approach significantly outperforms the state-of-the-arts in open-set domain adaptation.} }
Endnote
%0 Conference Paper %T Progressive Graph Learning for Open-Set Domain Adaptation %A Yadan Luo %A Zijian Wang %A Zi Huang %A Mahsa Baktashmotlagh %B Proceedings of the 37th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2020 %E Hal Daumé III %E Aarti Singh %F pmlr-v119-luo20b %I PMLR %P 6468--6478 %U https://proceedings.mlr.press/v119/luo20b.html %V 119 %X Domain shift is a fundamental problem in visual recognition which typically arises when the source and target data follow different distributions. The existing domain adaptation approaches which tackle this problem work in the "closed-set" setting with the assumption that the source and the target data share exactly the same classes of objects. In this paper, we tackle a more realistic problem of the "open-set" domain shift where the target data contains additional classes that were not present in the source data. More specifically, we introduce an end-to-end Progressive Graph Learning (PGL) framework where a graph neural network with episodic training is integrated to suppress underlying conditional shift and adversarial learning is adopted to close the gap between the source and target distributions. Compared to the existing open-set adaptation approaches, our approach guarantees to achieve a tighter upper bound of the target error. Extensive experiments on three standard open-set benchmarks evidence that our approach significantly outperforms the state-of-the-arts in open-set domain adaptation.
APA
Luo, Y., Wang, Z., Huang, Z. & Baktashmotlagh, M.. (2020). Progressive Graph Learning for Open-Set Domain Adaptation. Proceedings of the 37th International Conference on Machine Learning, in Proceedings of Machine Learning Research 119:6468-6478 Available from https://proceedings.mlr.press/v119/luo20b.html.

Related Material