ELSA: Efficient Label Shift Adaptation through the Lens of Semiparametric Models

Qinglong Tian, Xin Zhang, Jiwei Zhao
Proceedings of the 40th International Conference on Machine Learning, PMLR 202:34120-34142, 2023.

Abstract

We study the domain adaptation problem with label shift in this work. Under the label shift context, the marginal distribution of the label varies across the training and testing datasets, while the conditional distribution of features given the label is the same. Traditional label shift adaptation methods either suffer from large estimation errors or require cumbersome post-prediction calibrations. To address these issues, we first propose a moment-matching framework for adapting the label shift based on the geometry of the influence function. Under such a framework, we propose a novel method named $\underline{\mathrm{E}}$fficient $\underline{\mathrm{L}}$abel $\underline{\mathrm{S}}$hift $\underline{\mathrm{A}}$daptation (ELSA), in which the adaptation weights can be estimated by solving linear systems. Theoretically, the ELSA estimator is $\sqrt{n}$-consistent ($n$ is the sample size of the source data) and asymptotically normal. Empirically, we show that ELSA can achieve state-of-the-art estimation performances without post-prediction calibrations, thus, gaining computational efficiency.

Cite this Paper

BibTeX
@InProceedings{pmlr-v202-tian23a, title = {{ELSA}: Efficient Label Shift Adaptation through the Lens of Semiparametric Models}, author = {Tian, Qinglong and Zhang, Xin and Zhao, Jiwei}, booktitle = {Proceedings of the 40th International Conference on Machine Learning}, pages = {34120--34142}, year = {2023}, editor = {Krause, Andreas and Brunskill, Emma and Cho, Kyunghyun and Engelhardt, Barbara and Sabato, Sivan and Scarlett, Jonathan}, volume = {202}, series = {Proceedings of Machine Learning Research}, month = {23--29 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v202/tian23a/tian23a.pdf}, url = {https://proceedings.mlr.press/v202/tian23a.html}, abstract = {We study the domain adaptation problem with label shift in this work. Under the label shift context, the marginal distribution of the label varies across the training and testing datasets, while the conditional distribution of features given the label is the same. Traditional label shift adaptation methods either suffer from large estimation errors or require cumbersome post-prediction calibrations. To address these issues, we first propose a moment-matching framework for adapting the label shift based on the geometry of the influence function. Under such a framework, we propose a novel method named $\underline{\mathrm{E}}$fficient $\underline{\mathrm{L}}$abel $\underline{\mathrm{S}}$hift $\underline{\mathrm{A}}$daptation (ELSA), in which the adaptation weights can be estimated by solving linear systems. Theoretically, the ELSA estimator is $\sqrt{n}$-consistent ($n$ is the sample size of the source data) and asymptotically normal. Empirically, we show that ELSA can achieve state-of-the-art estimation performances without post-prediction calibrations, thus, gaining computational efficiency.} }
Endnote
%0 Conference Paper %T ELSA: Efficient Label Shift Adaptation through the Lens of Semiparametric Models %A Qinglong Tian %A Xin Zhang %A Jiwei Zhao %B Proceedings of the 40th International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2023 %E Andreas Krause %E Emma Brunskill %E Kyunghyun Cho %E Barbara Engelhardt %E Sivan Sabato %E Jonathan Scarlett %F pmlr-v202-tian23a %I PMLR %P 34120--34142 %U https://proceedings.mlr.press/v202/tian23a.html %V 202 %X We study the domain adaptation problem with label shift in this work. Under the label shift context, the marginal distribution of the label varies across the training and testing datasets, while the conditional distribution of features given the label is the same. Traditional label shift adaptation methods either suffer from large estimation errors or require cumbersome post-prediction calibrations. To address these issues, we first propose a moment-matching framework for adapting the label shift based on the geometry of the influence function. Under such a framework, we propose a novel method named $\underline{\mathrm{E}}$fficient $\underline{\mathrm{L}}$abel $\underline{\mathrm{S}}$hift $\underline{\mathrm{A}}$daptation (ELSA), in which the adaptation weights can be estimated by solving linear systems. Theoretically, the ELSA estimator is $\sqrt{n}$-consistent ($n$ is the sample size of the source data) and asymptotically normal. Empirically, we show that ELSA can achieve state-of-the-art estimation performances without post-prediction calibrations, thus, gaining computational efficiency.
APA
Tian, Q., Zhang, X. & Zhao, J.. (2023). ELSA: Efficient Label Shift Adaptation through the Lens of Semiparametric Models. Proceedings of the 40th International Conference on Machine Learning, in Proceedings of Machine Learning Research 202:34120-34142 Available from https://proceedings.mlr.press/v202/tian23a.html.

Related Material