Research map / Bandits and online learning / Multi-armed bandits
When Privacy Meets Partial Information: A Refined Analysis of Differentially Private Bandits
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 |
|---|---|---|---|
| Use the AdaP-KL-UCB framework instead of competing ε-global DP bandit algorithms | Bernoulli 5-arm bandit with means {0.75,0.625,0.5,0.375,0.25}, T = 10^7 · Bernoulli 5-arm bandit | average regret: 10 times lower regret | Helped |
| Increase ε across the high-privacy to low-privacy regimes | AdaP-KL-UCB · Bernoulli bandit with means {0.8, 0.1, 0.1, 0.1, 0.1}, T = 10^7 | regret: ε < 0.3; ε > 0.3 | Mixed |
| Use the AdaP-KL-UCB index rather than the AdaP-UCB index | Proposed adaptive-episode DP framework · Stochastic bandits | problem-dependent regret bound: d(μ_a, μ*) versus Δ_a^2 | Helped |
| Add Laplace noise to reward empirical means, compute private statistics from the last active episode, and use adaptive doubling episodes | Index-based optimistic bandit algorithms · Stochastic bandits | regret | Helped |
| Consider global privacy together with reward-distribution distinguishability | ε-global DP stochastic bandits · Stochastic bandits | problem-dependent regret lower bound: min(d_inf, 6 ε t_inf) | Mixed |
Closest papers by meaning
- Shuffle and Joint Differential Privacy for Generalized Linear Contextual Bandits (NeurIPS 2026 · similarity 0.90)
- Connecting Thompson Sampling and UCB: Towards More Efficient Trade-offs Between Privacy and Regret (ICML 2025 · similarity 0.90)
- Locally Private and Robust Multi-Armed Bandits (NeurIPS 2024 · similarity 0.90)
- Faster Rates for Private Adversarial Bandits (ICML 2025 · similarity 0.89)
- Fixed-Budget Differentially Private Best Arm Identification (ICLR 2024 · similarity 0.89)
- On Differentially Private Federated Linear Contextual Bandits (ICLR 2024 · similarity 0.89)
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.