Optimistic bandit convex optimization

Mehryar Mohri, Scott Yang

Research output: Contribution to journalArticle

Abstract

We introduce the general and powerful scheme of predicting information re-use in optimization algorithms. This allows us to devise a computationally efficient algorithm for bandit convex optimization with new state-of-the-art guarantees for both Lipschitz loss functions and loss functions with Lipschitz gradients. This is the first algorithm admitting both a polynomial time complexity and a regret that is polynomial in the dimension of the action space that improves upon the original regret bound for Lipschitz loss functions, achieving a regret of Õ(T11/16d3/8). Our algorithm further improves upon the best existing polynomial-in-dimension bound (both computationally and in terms of regret) for loss functions with Lipschitz gradients, achieving a regret of Õ (T8/13d5/3).

Original languageEnglish (US)
Pages (from-to)2297-2305
Number of pages9
JournalAdvances in Neural Information Processing Systems
StatePublished - 2016

Fingerprint

Convex optimization
Polynomials

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Information Systems
  • Signal Processing

Cite this

Optimistic bandit convex optimization. / Mohri, Mehryar; Yang, Scott.

In: Advances in Neural Information Processing Systems, 2016, p. 2297-2305.

Research output: Contribution to journalArticle

@article{63f5169c3fd541828bb6885edc031676,
title = "Optimistic bandit convex optimization",
abstract = "We introduce the general and powerful scheme of predicting information re-use in optimization algorithms. This allows us to devise a computationally efficient algorithm for bandit convex optimization with new state-of-the-art guarantees for both Lipschitz loss functions and loss functions with Lipschitz gradients. This is the first algorithm admitting both a polynomial time complexity and a regret that is polynomial in the dimension of the action space that improves upon the original regret bound for Lipschitz loss functions, achieving a regret of {\~O}(T11/16d3/8). Our algorithm further improves upon the best existing polynomial-in-dimension bound (both computationally and in terms of regret) for loss functions with Lipschitz gradients, achieving a regret of {\~O} (T8/13d5/3).",
author = "Mehryar Mohri and Scott Yang",
year = "2016",
language = "English (US)",
pages = "2297--2305",
journal = "Advances in Neural Information Processing Systems",
issn = "1049-5258",

}

TY - JOUR

T1 - Optimistic bandit convex optimization

AU - Mohri, Mehryar

AU - Yang, Scott

PY - 2016

Y1 - 2016

N2 - We introduce the general and powerful scheme of predicting information re-use in optimization algorithms. This allows us to devise a computationally efficient algorithm for bandit convex optimization with new state-of-the-art guarantees for both Lipschitz loss functions and loss functions with Lipschitz gradients. This is the first algorithm admitting both a polynomial time complexity and a regret that is polynomial in the dimension of the action space that improves upon the original regret bound for Lipschitz loss functions, achieving a regret of Õ(T11/16d3/8). Our algorithm further improves upon the best existing polynomial-in-dimension bound (both computationally and in terms of regret) for loss functions with Lipschitz gradients, achieving a regret of Õ (T8/13d5/3).

AB - We introduce the general and powerful scheme of predicting information re-use in optimization algorithms. This allows us to devise a computationally efficient algorithm for bandit convex optimization with new state-of-the-art guarantees for both Lipschitz loss functions and loss functions with Lipschitz gradients. This is the first algorithm admitting both a polynomial time complexity and a regret that is polynomial in the dimension of the action space that improves upon the original regret bound for Lipschitz loss functions, achieving a regret of Õ(T11/16d3/8). Our algorithm further improves upon the best existing polynomial-in-dimension bound (both computationally and in terms of regret) for loss functions with Lipschitz gradients, achieving a regret of Õ (T8/13d5/3).

UR - http://www.scopus.com/inward/record.url?scp=85018914432&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85018914432&partnerID=8YFLogxK

M3 - Article

SP - 2297

EP - 2305

JO - Advances in Neural Information Processing Systems

JF - Advances in Neural Information Processing Systems

SN - 1049-5258

ER -