Loss-Driven Bayesian Active Learning

Zhuoyue Huang, Freddie Bickford Smith, Tom Rainforth
Proceedings of The 29th International Conference on Artificial Intelligence and Statistics, PMLR 300:5140-5148, 2026.

Abstract

The central goal of active learning is to gather data that maximises downstream predictive performance, but popular approaches have limited flexibility in customising this data acquisition to different downstream problems and losses. We propose a rigorous loss-driven approach to Bayesian active learning that allows data acquisition to directly target the loss associated with a given decision problem. In particular, we show how any loss can be used to derive a unique objective for optimal data acquisition. Critically, we then show that any loss taking the form of a weighted Bregman divergence permits analytic computation of a central component of its corresponding objective, making the approach applicable in practice. In regression and classification experiments with a range of different losses, we find our approach reduces test losses relative to existing techniques.

Cite this Paper

BibTeX
@InProceedings{pmlr-v300-huang26b, title = { Loss-Driven Bayesian Active Learning }, author = {Huang, Zhuoyue and Smith, Freddie Bickford and Rainforth, Tom}, booktitle = {Proceedings of The 29th International Conference on Artificial Intelligence and Statistics}, pages = {5140--5148}, year = {2026}, editor = {Khan, Emtiyaz and Li, Yingzhen and Solin, Arno and Ramdas, Aaditya}, volume = {300}, series = {Proceedings of Machine Learning Research}, month = {02--05 May}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v300/main/assets/huang26b/huang26b.pdf}, url = {https://proceedings.mlr.press/v300/huang26b.html}, abstract = { The central goal of active learning is to gather data that maximises downstream predictive performance, but popular approaches have limited flexibility in customising this data acquisition to different downstream problems and losses. We propose a rigorous loss-driven approach to Bayesian active learning that allows data acquisition to directly target the loss associated with a given decision problem. In particular, we show how any loss can be used to derive a unique objective for optimal data acquisition. Critically, we then show that any loss taking the form of a weighted Bregman divergence permits analytic computation of a central component of its corresponding objective, making the approach applicable in practice. In regression and classification experiments with a range of different losses, we find our approach reduces test losses relative to existing techniques. } }
Endnote
%0 Conference Paper %T Loss-Driven Bayesian Active Learning %A Zhuoyue Huang %A Freddie Bickford Smith %A Tom Rainforth %B Proceedings of The 29th International Conference on Artificial Intelligence and Statistics %C Proceedings of Machine Learning Research %D 2026 %E Emtiyaz Khan %E Yingzhen Li %E Arno Solin %E Aaditya Ramdas %F pmlr-v300-huang26b %I PMLR %P 5140--5148 %U https://proceedings.mlr.press/v300/huang26b.html %V 300 %X The central goal of active learning is to gather data that maximises downstream predictive performance, but popular approaches have limited flexibility in customising this data acquisition to different downstream problems and losses. We propose a rigorous loss-driven approach to Bayesian active learning that allows data acquisition to directly target the loss associated with a given decision problem. In particular, we show how any loss can be used to derive a unique objective for optimal data acquisition. Critically, we then show that any loss taking the form of a weighted Bregman divergence permits analytic computation of a central component of its corresponding objective, making the approach applicable in practice. In regression and classification experiments with a range of different losses, we find our approach reduces test losses relative to existing techniques.
APA
Huang, Z., Smith, F.B. & Rainforth, T.. (2026). Loss-Driven Bayesian Active Learning . Proceedings of The 29th International Conference on Artificial Intelligence and Statistics, in Proceedings of Machine Learning Research 300:5140-5148 Available from https://proceedings.mlr.press/v300/huang26b.html.

Related Material