Optimal Learning from Label Proportions with General Loss Functions

Lorne Applebaum, Travis Dick, Claudio Gentile, Haim Kaplan, Tomer Koren
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:3228-3275, 2026.

Abstract

Motivated by problems in online advertising, we address the task of Learning from Label Proportions (LLP). We introduce a novel and versatile low-variance debiasing methodology to learn from aggregate label information, significantly advancing the state of the art in LLP. Our debiasing approach exhibits remarkable flexibility, seamlessly accommodating a broad spectrum of practically relevant loss functions across both binary and multi-class classification settings. By carefully combining our estimators with standard techniques, we improve sample complexity guarantees for a large class of losses of practical relevance. We also empirically validate the efficacy of our proposed approach across a diverse array of benchmark datasets, demonstrating compelling empirical advantages over standard baselines.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-applebaum26a, title = {Optimal Learning from Label Proportions with General Loss Functions}, author = {Applebaum, Lorne and Dick, Travis and Gentile, Claudio and Kaplan, Haim and Koren, Tomer}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {3228--3275}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/applebaum26a/applebaum26a.pdf}, url = {https://proceedings.mlr.press/v306/applebaum26a.html}, abstract = {Motivated by problems in online advertising, we address the task of Learning from Label Proportions (LLP). We introduce a novel and versatile low-variance debiasing methodology to learn from aggregate label information, significantly advancing the state of the art in LLP. Our debiasing approach exhibits remarkable flexibility, seamlessly accommodating a broad spectrum of practically relevant loss functions across both binary and multi-class classification settings. By carefully combining our estimators with standard techniques, we improve sample complexity guarantees for a large class of losses of practical relevance. We also empirically validate the efficacy of our proposed approach across a diverse array of benchmark datasets, demonstrating compelling empirical advantages over standard baselines.} }
Endnote
%0 Conference Paper %T Optimal Learning from Label Proportions with General Loss Functions %A Lorne Applebaum %A Travis Dick %A Claudio Gentile %A Haim Kaplan %A Tomer Koren %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-applebaum26a %I PMLR %P 3228--3275 %U https://proceedings.mlr.press/v306/applebaum26a.html %V 306 %X Motivated by problems in online advertising, we address the task of Learning from Label Proportions (LLP). We introduce a novel and versatile low-variance debiasing methodology to learn from aggregate label information, significantly advancing the state of the art in LLP. Our debiasing approach exhibits remarkable flexibility, seamlessly accommodating a broad spectrum of practically relevant loss functions across both binary and multi-class classification settings. By carefully combining our estimators with standard techniques, we improve sample complexity guarantees for a large class of losses of practical relevance. We also empirically validate the efficacy of our proposed approach across a diverse array of benchmark datasets, demonstrating compelling empirical advantages over standard baselines.
APA
Applebaum, L., Dick, T., Gentile, C., Kaplan, H. & Koren, T.. (2026). Optimal Learning from Label Proportions with General Loss Functions. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:3228-3275 Available from https://proceedings.mlr.press/v306/applebaum26a.html.

Related Material