Kullback-Leibler upper confidence bounds for optimal sequential allocation
From MaRDI portal
(Redirected from Publication:366995)
Abstract: We consider optimal sequential allocation in the context of the so-called stochastic multi-armed bandit model. We describe a generic index policy, in the sense of Gittins [J. R. Stat. Soc. Ser. B Stat. Methodol. 41 (1979) 148-177], based on upper confidence bounds of the arm payoffs computed using the Kullback-Leibler divergence. We consider two classes of distributions for which instances of this general idea are analyzed: the kl-UCB algorithm is designed for one-parameter exponential families and the empirical KL-UCB algorithm for bounded and finitely supported distributions. Our main contribution is a unified finite-time analysis of the regret of these algorithms that asymptotically matches the lower bounds of Lai and Robbins [Adv. in Appl. Math. 6 (1985) 4-22] and Burnetas and Katehakis [Adv. in Appl. Math. 17 (1996) 122-142], respectively. We also investigate the behavior of these algorithms when used with general bounded rewards, showing in particular that they provide significant improvements over the state-of-the-art.
Recommendations
- A minimax and asymptotically optimal algorithm for stochastic bandits
- Sample mean based index policies by O(log n) regret for the multi-armed bandit problem
- On Bayesian index policies for sequential resource allocation
- Adaptive treatment allocation and the multi-armed bandit problem
- Tuning Bandit Algorithms in Stochastic Environments
Cites work
- scientific article; zbMATH DE number 3687126 (Why is no real title available?)
- scientific article; zbMATH DE number 193660 (Why is no real title available?)
- scientific article; zbMATH DE number 3638998 (Why is no real title available?)
- scientific article; zbMATH DE number 1220667 (Why is no real title available?)
- scientific article; zbMATH DE number 1158743 (Why is no real title available?)
- ASYMPTOTIC BAYES ANALYSIS FOR THE FINITE-HORIZON ONE-ARMED-BANDIT PROBLEM
- An asymptotically optimal policy for finite support models in the multiarmed bandit problem
- Asymptotic Statistics
- Asymptotically efficient adaptive allocation rules
- Concentration inequalities and model selection. Ecole d'Eté de Probabilités de Saint-Flour XXXIII -- 2003.
- Empirical likelihood
- Exploration-exploitation tradeoff using variance estimates in multi-armed bandits
- Finite-time analysis of the multiarmed bandit problem
- Graphical models, exponential families, and variational inference
- Kullback-Leibler upper confidence bounds for optimal sequential allocation
- Multi-armed bandit allocation indices. With a foreword by Peter Whittle.
- On the Gittins index for multiarmed bandits
- On the Theory of Apportionment
- Optimal Adaptive Policies for Markov Decision Processes
- Optimal adaptive policies for sequential allocation problems
- Optimal stopping and dynamic allocation
- Probability Inequalities for Sums of Bounded Random Variables
- Regret analysis of stochastic and nonstochastic multi-armed bandit problems
- Regret bounds and minimax policies under partial monitoring
- Sample mean based index policies by O(log n) regret for the multi-armed bandit problem
- Sequential Tests of Statistical Hypotheses
- Some aspects of the sequential design of experiments
- Thompson sampling: an asymptotically optimal finite-time analysis
Cited in
(40)- Local Dvoretzky-Kiefer-Wolfowitz confidence bands
- Probabilistic learning constrained by realizations using a weak formulation of Fourier transform of probability measures
- A review of recent advances in empirical likelihood
- Asymptotically optimal multi-armed bandit policies under a cost constraint
- Lower bounds of success probabilities for high-fidelity approach in KLM scheme
- Probabilistic learning inference of boundary value problem with uncertainties based on Kullback-Leibler divergence under implicit constraints
- Learning the distribution with largest mean: two bandit frameworks
- Adaptive policies for perimeter surveillance problems
- Physics-constrained non-Gaussian probabilistic learning on manifolds
- Satisficing in Time-Sensitive Bandit Learning
- Batched bandit problems
- Learning to optimize via information-directed sampling
- Dealing with expert bias in collective decision-making
- Good arm identification via bandit feedback
- Explore first, exploit next: the true shape of regret in bandit problems
- Nonasymptotic sequential tests for overlapping hypotheses applied to near-optimal arm identification in bandit models
- scientific article; zbMATH DE number 6982311 (Why is no real title available?)
- Learning to optimize via posterior sampling
- Boundary crossing for general exponential families
- Infinite Arms Bandit: Optimality via Confidence Bounds
- Scalar utility theory and proportional processing: what does it actually imply?
- Boundary crossing probabilities for general exponential families
- Infomax strategies for an optimal balance between exploration and exploitation
- A confirmation of a conjecture on Feldman’s two-armed bandit problem
- Bayesian adaptive bandit-based designs using the Gittins index for multi-armed trials with normally distributed endpoints
- scientific article; zbMATH DE number 7370545 (Why is no real title available?)
- A minimax and asymptotically optimal algorithm for stochastic bandits
- Asymptotically optimal algorithms for budgeted multiple play bandits
- Normal bandits of unknown means and variances
- scientific article; zbMATH DE number 7370531 (Why is no real title available?)
- scientific article; zbMATH DE number 7370524 (Why is no real title available?)
- Technical note -- A note on the equivalence of upper confidence bounds and Gittins indices for patient agents
- Finite-time analysis for the knowledge-gradient policy
- scientific article; zbMATH DE number 7626761 (Why is no real title available?)
- On Bayesian index policies for sequential resource allocation
- A better resource allocation algorithm with semi-bandit feedback
- Kullback-Leibler upper confidence bounds for optimal sequential allocation
- Tuning Bandit Algorithms in Stochastic Environments
- Regret bounds for Narendra-Shapiro bandit algorithms
- The multi-armed bandit problem: an efficient nonparametric solution
This page was built for publication: Kullback-Leibler upper confidence bounds for optimal sequential allocation
Report a bug (only for logged in users!)Click here to report a bug for this page (MaRDI item Q366995)
