Research map / Bandits and online learning / Multi-armed bandits
On Limited-Memory Subsampling Strategies for Bandits
Dorian Baudry, Yoan Russac, Olivier Cappé
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 changed | Where | Measured | Effect |
|---|---|---|---|
| Use deterministic last-block subsampling for duels | LB-SDA · Stationary one-parameter exponential-family bandits | regret: \cR(T) \leq \sum_{k: \mu_k < \mu^\star} \frac{1+\epsilon}{\kl(\mu_k, \mu^\star)} \log(T) + C(\nu, \epsilon) | Improved |
| Limit stored observations per arm using m_r with m_r/\log(r) \to \infty (ablation) | LB-SDA · Stationary one-parameter exponential-family bandits | regret: \cR_T \leq \sum_{k: \mu_k < \mu^\star} \frac{1+\epsilon}{\kl(\mu_k, \mu^\star)} \log(T) + C'(\nu, \epsilon, \cM) | No clear effect |
| Limit memory to m_r = \log(r)^2 + 50 (ablation) | LB-SDA · Stationary Bernoulli bandit, K=2, T=10000 | regret | Hurt |
| Use SW-LB-SDA instead of competing methods | Abruptly changing Bernoulli bandit · Bernoulli instance with three arms | regret | Improved |
| Use SW-LB-SDA instead of competing methods | Abruptly changing Gaussian bandit · Gaussian instance with three arms and fixed standard deviation | regret | Improved |
| Use SW-LB-SDA without supplying variance information | Non-stationary bandit algorithms · Gaussian bandit with changing standard errors | regret | Improved |
Closest papers by meaning
- Almost Free: Self-concordance in Natural Exponential Families and an Application to Bandits (NeurIPS 2024 · similarity 0.88)
- Local Anti-Concentration Class: Logarithmic Regret for Greedy Linear Contextual Bandit (NeurIPS 2024 · similarity 0.88)
- Non-stationary Bandit Convex Optimization: A Comprehensive Study (NeurIPS 2025 · similarity 0.88)
- Active Context Selection Improves Simple Regret in Contextual Bandits (NeurIPS 2026 · similarity 0.88)
- Logistic Bandits with $\tilde{O}(\sqrt{dT})$ Regret without Context Diversity Assumptions (NeurIPS 2026 · similarity 0.88)
- Block Optimism for Nonstationary Bandits with Latent Linear Dynamics (NeurIPS 2026 · 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.
