AI Paper Atlas

Research map / Bandits and online learning / Online convex optimization and games

Multi-Armed Bandits with Metric Movement Costs

Tomer Koren, Roi Livni, Yishay Mansour

NeurIPS 2017 · Bandits and online learning › Online convex optimization and games · approach: Best-of-both-worlds bandit algorithms · 5 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
Use SMB with a tree metric dominating the finite action metricFinite-metric bandits with movement costs · Finite metric spacesmovement regret: \wt{O}(\max\{\dimu^{1/3} T^{2/3}, \sqrt{kT}\})Improved
Allow arbitrary finite metric movement costs rather than only uniform switching costsFinite-metric bandits · Finite metric spacesmovement regret: \wt{O}(\max\{\dimu^{1/3} T^{2/3}, \sqrt{kT}\})Mixed
Use SMB directly on a general HST instead of only a complete binary balanced HSTSMB · 2-HST metric spacesmovement regret: O\LR{ H \sqrt{2^H T \mathcal{C} \smash{\log\mathcal{C} }} + H 2^{-H} T }Improved
Discretize the infinite metric space and use the finite-space algorithm under Lipschitz lossesBandits with movement costs · Infinite metric spaces with 1-Lipschitz lossesmovement regret: \wt{O}(T^\frac{d+1}{d+2})Improved
Add movement costs to the Lipschitz metric-space bandit settingLipschitz metric-space bandits · Infinite metric spaces with matching upper and lower Minkowski dimensionsregret: \wt{O}(T^\frac{d+1}{d+2})No clear effect

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy