Stochastic Constrained Contextual Bandits via Lyapunov Optimization Based Estimation to Decision Framework

Hengquan Guo, Xin Liu
Proceedings of Thirty Seventh Conference on Learning Theory, PMLR 247:2204-2231, 2024.

Abstract

This paper studies the problem of stochastic constrained contextual bandits (CCB) under general realizability condition where the expected rewards and costs are within general function classes. We propose LOE2D, a Lyapunov Optimization Based Estimation to Decision framework with online regression oracles for learning reward/constraint. LOE2D establishes \TildeO(T34U14) regret and constraint violation, which can be further refined to \TildeO(min when the Slater condition holds in the underlying offline problem with the Slater “constant” \varepsilon=\Omega(\sqrt{U/T}), where U denotes the error bounds of online regression oracles. These results improve LagrangeCBwLC in two aspects: i) our results hold without any prior information while LagrangeCBwLC requires the knowledge of Slater constant to design a proper learning rate; ii) our results hold when \varepsilon=\Omega(\sqrt{U/T}) while LagrangeCBwLC requires a constant margin \varepsilon=\Omega(1). These improvements stem from two novel techniques: violation-adaptive learning in E2D module and multi-step Lyapunov drift analysis in bounding constraint violation. The experiments further justify LOE2D outperforms the baseline algorithm.

Cite this Paper

BibTeX
@InProceedings{pmlr-v247-guo24a, title = {Stochastic Constrained Contextual Bandits via Lyapunov Optimization Based Estimation to Decision Framework}, author = {Guo, Hengquan and Liu, Xin}, booktitle = {Proceedings of Thirty Seventh Conference on Learning Theory}, pages = {2204--2231}, year = {2024}, editor = {Agrawal, Shipra and Roth, Aaron}, volume = {247}, series = {Proceedings of Machine Learning Research}, month = {30 Jun--03 Jul}, publisher = {PMLR}, pdf = {https://proceedings.mlr.press/v247/guo24a/guo24a.pdf}, url = {https://proceedings.mlr.press/v247/guo24a.html}, abstract = {This paper studies the problem of stochastic constrained contextual bandits (CCB) under general realizability condition where the expected rewards and costs are within general function classes. We propose LOE2D, a Lyapunov Optimization Based Estimation to Decision framework with online regression oracles for learning reward/constraint. LOE2D establishes $\Tilde O(T^{\frac{3}{4}}U^{\frac{1}{4}})$ regret and constraint violation, which can be further refined to $\Tilde O(\min\{\sqrt{TU}/\varepsilon^2, T^{\frac{3}{4}}U^{\frac{1}{4}}\})$ when the Slater condition holds in the underlying offline problem with the Slater “constant” $ \varepsilon=\Omega(\sqrt{U/T}),$ where $U$ denotes the error bounds of online regression oracles. These results improve LagrangeCBwLC in two aspects: i) our results hold without any prior information while LagrangeCBwLC requires the knowledge of Slater constant to design a proper learning rate; ii) our results hold when $\varepsilon=\Omega(\sqrt{U/T})$ while LagrangeCBwLC requires a constant margin $\varepsilon=\Omega(1).$ These improvements stem from two novel techniques: violation-adaptive learning in E2D module and multi-step Lyapunov drift analysis in bounding constraint violation. The experiments further justify LOE2D outperforms the baseline algorithm. } }
Endnote
%0 Conference Paper %T Stochastic Constrained Contextual Bandits via Lyapunov Optimization Based Estimation to Decision Framework %A Hengquan Guo %A Xin Liu %B Proceedings of Thirty Seventh Conference on Learning Theory %C Proceedings of Machine Learning Research %D 2024 %E Shipra Agrawal %E Aaron Roth %F pmlr-v247-guo24a %I PMLR %P 2204--2231 %U https://proceedings.mlr.press/v247/guo24a.html %V 247 %X This paper studies the problem of stochastic constrained contextual bandits (CCB) under general realizability condition where the expected rewards and costs are within general function classes. We propose LOE2D, a Lyapunov Optimization Based Estimation to Decision framework with online regression oracles for learning reward/constraint. LOE2D establishes $\Tilde O(T^{\frac{3}{4}}U^{\frac{1}{4}})$ regret and constraint violation, which can be further refined to $\Tilde O(\min\{\sqrt{TU}/\varepsilon^2, T^{\frac{3}{4}}U^{\frac{1}{4}}\})$ when the Slater condition holds in the underlying offline problem with the Slater “constant” $ \varepsilon=\Omega(\sqrt{U/T}),$ where $U$ denotes the error bounds of online regression oracles. These results improve LagrangeCBwLC in two aspects: i) our results hold without any prior information while LagrangeCBwLC requires the knowledge of Slater constant to design a proper learning rate; ii) our results hold when $\varepsilon=\Omega(\sqrt{U/T})$ while LagrangeCBwLC requires a constant margin $\varepsilon=\Omega(1).$ These improvements stem from two novel techniques: violation-adaptive learning in E2D module and multi-step Lyapunov drift analysis in bounding constraint violation. The experiments further justify LOE2D outperforms the baseline algorithm.
APA
Guo, H. & Liu, X.. (2024). Stochastic Constrained Contextual Bandits via Lyapunov Optimization Based Estimation to Decision Framework. Proceedings of Thirty Seventh Conference on Learning Theory, in Proceedings of Machine Learning Research 247:2204-2231 Available from https://proceedings.mlr.press/v247/guo24a.html.

Related Material