DAL: A Practical Prior-Free Black-Box Framework for Piecewise Stationary Bandits

Argyrios Gerogiannis, Yu-Han Huang, Subhonmesh Bose, Venugopal Veeravalli
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:34611-34638, 2026.

Abstract

We introduce a practical, black-box framework termed Detection Augmented Learning (DAL) for the problem of piecewise stationary bandits without knowledge of the underlying non-stationarity. DAL accepts any stationary bandit algorithm with order-optimal regret as input and augments it with a change detector, enabling applicability to all common bandit variants. Extensive experimentation demonstrates that DAL consistently surpasses all state-of-the-art methods across diverse non-stationary scenarios, including synthetic benchmarks and real-world datasets, underscoring its versatility and scalability. We provide theoretical insights into DAL’s strong empirical performance, complemented by thorough empirical validation.

Cite this Paper

BibTeX
@InProceedings{pmlr-v306-gerogiannis26a, title = {{DAL}: A Practical Prior-Free Black-Box Framework for Piecewise Stationary Bandits}, author = {Gerogiannis, Argyrios and Huang, Yu-Han and Bose, Subhonmesh and Veeravalli, Venugopal}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {34611--34638}, 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/gerogiannis26a/gerogiannis26a.pdf}, url = {https://proceedings.mlr.press/v306/gerogiannis26a.html}, abstract = {We introduce a practical, black-box framework termed Detection Augmented Learning (DAL) for the problem of piecewise stationary bandits without knowledge of the underlying non-stationarity. DAL accepts any stationary bandit algorithm with order-optimal regret as input and augments it with a change detector, enabling applicability to all common bandit variants. Extensive experimentation demonstrates that DAL consistently surpasses all state-of-the-art methods across diverse non-stationary scenarios, including synthetic benchmarks and real-world datasets, underscoring its versatility and scalability. We provide theoretical insights into DAL’s strong empirical performance, complemented by thorough empirical validation.} }
Endnote
%0 Conference Paper %T DAL: A Practical Prior-Free Black-Box Framework for Piecewise Stationary Bandits %A Argyrios Gerogiannis %A Yu-Han Huang %A Subhonmesh Bose %A Venugopal Veeravalli %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-gerogiannis26a %I PMLR %P 34611--34638 %U https://proceedings.mlr.press/v306/gerogiannis26a.html %V 306 %X We introduce a practical, black-box framework termed Detection Augmented Learning (DAL) for the problem of piecewise stationary bandits without knowledge of the underlying non-stationarity. DAL accepts any stationary bandit algorithm with order-optimal regret as input and augments it with a change detector, enabling applicability to all common bandit variants. Extensive experimentation demonstrates that DAL consistently surpasses all state-of-the-art methods across diverse non-stationary scenarios, including synthetic benchmarks and real-world datasets, underscoring its versatility and scalability. We provide theoretical insights into DAL’s strong empirical performance, complemented by thorough empirical validation.
APA
Gerogiannis, A., Huang, Y., Bose, S. & Veeravalli, V.. (2026). DAL: A Practical Prior-Free Black-Box Framework for Piecewise Stationary Bandits. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:34611-34638 Available from https://proceedings.mlr.press/v306/gerogiannis26a.html.

Related Material