An Improved Parametrization and Analysis of the EXP3++ Algorithm for Stochastic and Adversarial Bandits

Yevgeny Seldin, Gábor Lugosi
Proceedings of the 2017 Conference on Learning Theory, PMLR 65:1743-1759, 2017.

Abstract

We present a new strategy for gap estimation in randomized algorithms for multiarmed bandits and combine it with the EXP3++ algorithm of Seldin and Slivkins (2014). In the stochastic regime the strategy reduces dependence of regret on a time horizon from $(\ln t)^3$ to $(\ln t)^2$ and eliminates an additive factor of order $∆e^1/∆^2$, where $∆$ is the minimal gap of a problem instance. In the adversarial regime regret guarantee remains unchanged.

Cite this Paper

BibTeX
@InProceedings{pmlr-v65-seldin17a, title = {An Improved Parametrization and Analysis of the {EXP3++} Algorithm for Stochastic and Adversarial Bandits}, author = {Seldin, Yevgeny and Lugosi, Gábor}, booktitle = {Proceedings of the 2017 Conference on Learning Theory}, pages = {1743--1759}, year = {2017}, editor = {Kale, Satyen and Shamir, Ohad}, volume = {65}, series = {Proceedings of Machine Learning Research}, month = {07--10 Jul}, publisher = {PMLR}, pdf = {http://proceedings.mlr.press/v65/seldin17a/seldin17a.pdf}, url = {https://proceedings.mlr.press/v65/seldin17a.html}, abstract = {We present a new strategy for gap estimation in randomized algorithms for multiarmed bandits and combine it with the EXP3++ algorithm of Seldin and Slivkins (2014). In the stochastic regime the strategy reduces dependence of regret on a time horizon from $(\ln t)^3$ to $(\ln t)^2$ and eliminates an additive factor of order $∆e^1/∆^2$, where $∆$ is the minimal gap of a problem instance. In the adversarial regime regret guarantee remains unchanged.} }
Endnote
%0 Conference Paper %T An Improved Parametrization and Analysis of the EXP3++ Algorithm for Stochastic and Adversarial Bandits %A Yevgeny Seldin %A Gábor Lugosi %B Proceedings of the 2017 Conference on Learning Theory %C Proceedings of Machine Learning Research %D 2017 %E Satyen Kale %E Ohad Shamir %F pmlr-v65-seldin17a %I PMLR %P 1743--1759 %U https://proceedings.mlr.press/v65/seldin17a.html %V 65 %X We present a new strategy for gap estimation in randomized algorithms for multiarmed bandits and combine it with the EXP3++ algorithm of Seldin and Slivkins (2014). In the stochastic regime the strategy reduces dependence of regret on a time horizon from $(\ln t)^3$ to $(\ln t)^2$ and eliminates an additive factor of order $∆e^1/∆^2$, where $∆$ is the minimal gap of a problem instance. In the adversarial regime regret guarantee remains unchanged.
APA
Seldin, Y. & Lugosi, G.. (2017). An Improved Parametrization and Analysis of the EXP3++ Algorithm for Stochastic and Adversarial Bandits. Proceedings of the 2017 Conference on Learning Theory, in Proceedings of Machine Learning Research 65:1743-1759 Available from https://proceedings.mlr.press/v65/seldin17a.html.

Related Material