Kernel-based methods for bandit convex optimization

@inproceedings{Bubeck2017KernelbasedMF,
  title={Kernel-based methods for bandit convex optimization},
  author={S{\'e}bastien Bubeck and Yin Tat Lee and Ronen Eldan},
  booktitle={STOC},
  year={2017}
}
We consider the adversarial convex bandit problem and we build the first <i>poly</i>(<i>T</i>)-time algorithm with <i>poly</i>(<i>n</i>) √<i>T</i>-regret for this problem. To do so we introduce three new ideas in the derivative-free optimization literature: (i) kernel methods, (ii) a generalization of Bernoulli convolutions, and (iii) a new annealing schedule for exponential weights (with increasing learning rate). The basic version of our algorithm achieves Õ(<i>n</i><sup>9.5</sup> √<i>T… CONTINUE READING
Highly Cited
This paper has 33 citations. REVIEW CITATIONS
Related Discussions
This paper has been referenced on Twitter 17 times. VIEW TWEETS

References

Publications referenced by this paper.

On an algorithm for the minimization of convex func

A. Levin
al Symposium on, • 2015
View 6 Excerpts
Highly Influenced

Similar Papers

Loading similar papers…