Research map / Bandits and online learning / Multi-armed bandits
Scalable Discrete Sampling as a Multi-Armed Bandit Problem
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 changed | Where | Measured | Effect |
|---|---|---|---|
| Adding a Racing-Normal (finite-population, Gaussian-assumption) subsampled bandit test to Gibbs for discrete variable sampling | Author coreference Gibbs sampling · Author Coreference (DBLP+Rexa) | convergence speed in factor evaluations: about three times faster than exact Gibbs | Helped |
| Subsampling discrete sampling with Racing-Normal in ARCH model selection Gibbs and MH | Bayesian ARCH model selection · Shanghai stock index ARCH model selection | effective sample size: over twice the effective sample size | Helped |
| Replacing concentration-bound (EBS) or union-bound G with the Racing-Normal bound using doubling schedule and first mini-batch 50 | Synthetic discrete sampling (D=10) · Synthetic data | number of rewards sampled to meet error tolerance | Helped |
| Using adapted lil'UCB instead of Racing-Normal bound | Synthetic discrete sampling · Synthetic data (D=100, Uniform) | error vs tolerance delta | Hurt |
| Using control variates (Taylor expansion of log-likelihood) to reduce reward variance (ablation) | Racing-Normal in ARCH model selection · Shanghai stock index ARCH model selection | reward variance: reduces reward variance by 2~3 orders of magnitude | Helped |
| Using the pairwise variance estimate instead of marginal estimate in Racing (ablation) | Synthetic discrete sampling · Synthetic data | error and reward sample count | Helped |
| Doubling mini-batch schedule (vs Bernstein-type schedules) for Racing-Normal (ablation) | Racing-Normal robustness with LogNormal rewards · Synthetic data | robustness across reward distributions | Helped |
Closest papers by meaning
- On Approximate Thompson Sampling with Langevin Algorithms (ICML 2020 · similarity 0.89)
- Thompson Sampling with Less Exploration is Fast and Optimal (ICML 2023 · similarity 0.89)
- Racing Thompson: an Efficient Algorithm for Thompson Sampling with Non-conjugate Priors (ICML 2018 · similarity 0.89)
- Semi-Parametric Sampling for Stochastic Bandits with Many Arms (AAAI 2019 · similarity 0.88)
- $\mathtt{VITS}$ : Variational Inference Thompson Sampling for contextual bandits (ICML 2024 · similarity 0.87)
- The Choice of Noninformative Priors for Thompson Sampling in Multiparameter Bandit Models (AAAI 2024 · similarity 0.87)
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.