Improved regret for zeroth-order adversarial bandit convex optimisation
Mathematical Statistics and Learning · 2020
DOI: 10.4171/msl/17 ↗Kurz gefasst
Fachartikel in der Zeitschrift Mathematical Statistics and Learning, 2020. Verfasst von Tor Lattimore. Die Arbeit wurde bisher 11-mal zitiert (Crossref).
Abstract (Originalsprache)
We prove that the information-theoretic upper bound on the minimax regret for zeroth-order adversarial bandit convex optimisation is at most O(d^{2.5} \sqrt{n} \mathrm {log}(n)) , where d is the dimension and n is the number of interactions. This improves on the bound of O(d^{9.5} \sqrt{n} \mathrm {log}(n)^{7.5}) by Bubeck et al. (2017). The proof is based on identifying an improved exploratory distribution for convex functions.
Zitieren
APA
Lattimore, T. (2020). Improved regret for zeroth-order adversarial bandit convex optimisation. Mathematical Statistics and Learning. https://doi.org/10.4171/msl/17
Vancouver
Lattimore T. Improved regret for zeroth-order adversarial bandit convex optimisation. Mathematical Statistics and Learning. 2020. doi:10.4171/msl/17
BibTeX
@article{Lattimore2020,
author = {Lattimore, Tor},
title = {Improved regret for zeroth-order adversarial bandit convex optimisation},
journal = {Mathematical Statistics and Learning},
year = {2020},
doi = {10.4171/msl/17}
}
EUVIMED