Current browse context:
math.ST
Change to browse by:
References & Citations
Mathematics > Statistics Theory
Title: Refined Lower Bounds for Adversarial Bandits
(Submitted on 24 May 2016 (v1), last revised 27 Feb 2017 (this version, v2))
Abstract: We provide new lower bounds on the regret that must be suffered by adversarial bandit algorithms. The new results show that recent upper bounds that either (a) hold with high-probability or (b) depend on the total lossof the best arm or (c) depend on the quadratic variation of the losses, are close to tight. Besides this we prove two impossibility results. First, the existence of a single arm that is optimal in every round cannot improve the regret in the worst case. Second, the regret cannot scale with the effective range of the losses. In contrast, both results are possible in the full-information setting.
Submission history
From: Sebastien Gerchinovitz [view email][v1] Tue, 24 May 2016 12:36:47 GMT (22kb)
[v2] Mon, 27 Feb 2017 13:48:10 GMT (23kb)
Link back to: arXiv, form interface, contact.