Connecting the Dots with Landmarks: Discriminatively Learning Domain-Invariant Features for Unsupervised Domain Adaptation

Boqing Gong, Kristen Grauman, Fei Sha
; Proceedings of the 30th International Conference on Machine Learning, PMLR 28(1):222-230, 2013.

Abstract

Learning domain-invariant features is of vital importance to unsupervised domain adaptation, where classifiers trained on the source domain need to be adapted to a different target domain for which no labeled examples are available. In this paper, we propose a novel approach for learning such features. The central idea is to exploit the existence of landmarks, which are a subset of labeled data instances in the source domain that are distributed most similarly to the target domain. Our approach automatically discovers the landmarks and use them to bridge the source to the target by constructing provably easier auxiliary domain adaptation tasks. The solutions of those auxiliary tasks form the basis to compose invariant features for the original task. We show how this composition can be optimized discriminatively without requiring labels from the target domain. We validate the method on standard benchmark datasets for visual object recognition and sentiment analysis of text. Empirical results show the proposed method outperforms the state-of-the-art significantly.

Cite this Paper


BibTeX
@InProceedings{pmlr-v28-gong13, title = {Connecting the Dots with Landmarks: Discriminatively Learning Domain-Invariant Features for Unsupervised Domain Adaptation}, author = {Boqing Gong and Kristen Grauman and Fei Sha}, booktitle = {Proceedings of the 30th International Conference on Machine Learning}, pages = {222--230}, year = {2013}, editor = {Sanjoy Dasgupta and David McAllester}, volume = {28}, number = {1}, series = {Proceedings of Machine Learning Research}, address = {Atlanta, Georgia, USA}, month = {17--19 Jun}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v28/gong13.pdf}, url = {http://proceedings.mlr.press/v28/gong13.html}, abstract = {Learning domain-invariant features is of vital importance to unsupervised domain adaptation, where classifiers trained on the source domain need to be adapted to a different target domain for which no labeled examples are available. In this paper, we propose a novel approach for learning such features. The central idea is to exploit the existence of landmarks, which are a subset of labeled data instances in the source domain that are distributed most similarly to the target domain. Our approach automatically discovers the landmarks and use them to bridge the source to the target by constructing provably easier auxiliary domain adaptation tasks. The solutions of those auxiliary tasks form the basis to compose invariant features for the original task. We show how this composition can be optimized discriminatively without requiring labels from the target domain. We validate the method on standard benchmark datasets for visual object recognition and sentiment analysis of text. Empirical results show the proposed method outperforms the state-of-the-art significantly. } }
Endnote
%0 Conference Paper %T Connecting the Dots with Landmarks: Discriminatively Learning Domain-Invariant Features for Unsupervised Domain Adaptation %A Boqing Gong %A Kristen Grauman %A Fei Sha %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-gong13 %I PMLR %J Proceedings of Machine Learning Research %P 222--230 %U http://proceedings.mlr.press %V 28 %N 1 %W PMLR %X Learning domain-invariant features is of vital importance to unsupervised domain adaptation, where classifiers trained on the source domain need to be adapted to a different target domain for which no labeled examples are available. In this paper, we propose a novel approach for learning such features. The central idea is to exploit the existence of landmarks, which are a subset of labeled data instances in the source domain that are distributed most similarly to the target domain. Our approach automatically discovers the landmarks and use them to bridge the source to the target by constructing provably easier auxiliary domain adaptation tasks. The solutions of those auxiliary tasks form the basis to compose invariant features for the original task. We show how this composition can be optimized discriminatively without requiring labels from the target domain. We validate the method on standard benchmark datasets for visual object recognition and sentiment analysis of text. Empirical results show the proposed method outperforms the state-of-the-art significantly.
RIS
TY - CPAPER TI - Connecting the Dots with Landmarks: Discriminatively Learning Domain-Invariant Features for Unsupervised Domain Adaptation AU - Boqing Gong AU - Kristen Grauman AU - Fei Sha BT - Proceedings of the 30th International Conference on Machine Learning PY - 2013/02/13 DA - 2013/02/13 ED - Sanjoy Dasgupta ED - David McAllester ID - pmlr-v28-gong13 PB - PMLR SP - 222 DP - PMLR EP - 230 L1 - http://proceedings.mlr.press/v28/gong13.pdf UR - http://proceedings.mlr.press/v28/gong13.html AB - Learning domain-invariant features is of vital importance to unsupervised domain adaptation, where classifiers trained on the source domain need to be adapted to a different target domain for which no labeled examples are available. In this paper, we propose a novel approach for learning such features. The central idea is to exploit the existence of landmarks, which are a subset of labeled data instances in the source domain that are distributed most similarly to the target domain. Our approach automatically discovers the landmarks and use them to bridge the source to the target by constructing provably easier auxiliary domain adaptation tasks. The solutions of those auxiliary tasks form the basis to compose invariant features for the original task. We show how this composition can be optimized discriminatively without requiring labels from the target domain. We validate the method on standard benchmark datasets for visual object recognition and sentiment analysis of text. Empirical results show the proposed method outperforms the state-of-the-art significantly. ER -
APA
Gong, B., Grauman, K. & Sha, F.. (2013). Connecting the Dots with Landmarks: Discriminatively Learning Domain-Invariant Features for Unsupervised Domain Adaptation. Proceedings of the 30th International Conference on Machine Learning, in PMLR 28(1):222-230

Related Material