Aggregating From Multiple Target-Shifted Sources

Changjian Shui, Zijian Li, Jiaqi Li, Christian Gagné, Charles X Ling, Boyu Wang
Proceedings of the 38th International Conference on Machine Learning, PMLR 139:9638-9648, 2021.

Abstract

Multi-source domain adaptation aims at leveraging the knowledge from multiple tasks for predicting a related target domain. Hence, a crucial aspect is to properly combine different sources based on their relations. In this paper, we analyzed the problem for aggregating source domains with different label distributions, where most recent source selection approaches fail. Our proposed algorithm differs from previous approaches in two key ways: the model aggregates multiple sources mainly through the similarity of semantic conditional distribution rather than marginal distribution; the model proposes a unified framework to select relevant sources for three popular scenarios, i.e., domain adaptation with limited label on target domain, unsupervised domain adaptation and label partial unsupervised domain adaption. We evaluate the proposed method through extensive experiments. The empirical results significantly outperform the baselines.

Cite this Paper

BibTeX
@InProceedings{pmlr-v139-shui21a, title = {Aggregating From Multiple Target-Shifted Sources}, author = {Shui, Changjian and Li, Zijian and Li, Jiaqi and Gagn{\'e}, Christian and Ling, Charles X and Wang, Boyu}, booktitle = {Proceedings of the 38th International Conference on Machine Learning}, pages = {9638--9648}, year = {2021}, editor = {Meila, Marina and Zhang, Tong}, volume = {139}, series = {Proceedings of Machine Learning Research}, month = {18--24 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v139/shui21a/shui21a.pdf}, url = {https://proceedings.mlr.press/v139/shui21a.html}, abstract = {Multi-source domain adaptation aims at leveraging the knowledge from multiple tasks for predicting a related target domain. Hence, a crucial aspect is to properly combine different sources based on their relations. In this paper, we analyzed the problem for aggregating source domains with different label distributions, where most recent source selection approaches fail. Our proposed algorithm differs from previous approaches in two key ways: the model aggregates multiple sources mainly through the similarity of semantic conditional distribution rather than marginal distribution; the model proposes a unified framework to select relevant sources for three popular scenarios, i.e., domain adaptation with limited label on target domain, unsupervised domain adaptation and label partial unsupervised domain adaption. We evaluate the proposed method through extensive experiments. The empirical results significantly outperform the baselines.} }
Endnote
%0 Conference Paper %T Aggregating From Multiple Target-Shifted Sources %A Changjian Shui %A Zijian Li %A Jiaqi Li %A Christian Gagné %A Charles X Ling %A Boyu Wang %B Proceedings of the 38th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2021 %E Marina Meila %E Tong Zhang %F pmlr-v139-shui21a %I PMLR %P 9638--9648 %U https://proceedings.mlr.press/v139/shui21a.html %V 139 %X Multi-source domain adaptation aims at leveraging the knowledge from multiple tasks for predicting a related target domain. Hence, a crucial aspect is to properly combine different sources based on their relations. In this paper, we analyzed the problem for aggregating source domains with different label distributions, where most recent source selection approaches fail. Our proposed algorithm differs from previous approaches in two key ways: the model aggregates multiple sources mainly through the similarity of semantic conditional distribution rather than marginal distribution; the model proposes a unified framework to select relevant sources for three popular scenarios, i.e., domain adaptation with limited label on target domain, unsupervised domain adaptation and label partial unsupervised domain adaption. We evaluate the proposed method through extensive experiments. The empirical results significantly outperform the baselines.
APA
Shui, C., Li, Z., Li, J., Gagné, C., Ling, C.X. & Wang, B.. (2021). Aggregating From Multiple Target-Shifted Sources. Proceedings of the 38th International Conference on Machine Learning, in Proceedings of Machine Learning Research 139:9638-9648 Available from https://proceedings.mlr.press/v139/shui21a.html.

Related Material