Contextual Bandits with Continuous Actions: Smoothing, Zooming, and Adapting

Abstract

We study contextual bandit learning for any competitor policy class and continuous action space. We obtain two qualitatively different regret bounds: one competes with a smoothed version of the policy class under no continuity assumptions, while the other requires standard Lipschitz assumptions. Both bounds exhibit data-dependent “zooming" behavior and, with no tuning, yield improved guarantees for benign problems. We also study adapting to unknown smoothness parameters, establishing a price-of-adaptivity and deriving optimal adaptive algorithms that require no additional information.

Cite

Text

Krishnamurthy et al. "Contextual Bandits with Continuous Actions: Smoothing, Zooming, and Adapting." Conference on Learning Theory, 2019.

Markdown

[Krishnamurthy et al. "Contextual Bandits with Continuous Actions: Smoothing, Zooming, and Adapting." Conference on Learning Theory, 2019.](https://mlanthology.org/colt/2019/krishnamurthy2019colt-contextual/)

BibTeX

@inproceedings{krishnamurthy2019colt-contextual,
  title     = {{Contextual Bandits with Continuous Actions: Smoothing, Zooming, and Adapting}},
  author    = {Krishnamurthy, Akshay and Langford, John and Slivkins, Aleksandrs and Zhang, Chicheng},
  booktitle = {Conference on Learning Theory},
  year      = {2019},
  pages     = {2025-2027},
  volume    = {99},
  url       = {https://mlanthology.org/colt/2019/krishnamurthy2019colt-contextual/}
}