Research map / Bandits and online learning / Contextual bandits and Thompson sampling
Learning Neural Contextual Bandits through Perturbed Rewards
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 perturbed rewards for exploration instead of explicit confidence-set exploration | Neural bandits · Synthetic nonlinear reward functions | cumulative regret | No clear effect |
| Use diagonal covariance approximation instead of the full covariance matrix | NeuralUCB and NeuralTS · Synthetic nonlinear reward functions | cumulative regret | Hurt |
| Use diagonal covariance approximation instead of the full covariance matrix | NeuralUCB and NeuralTS · Mushroom | cumulative regret | Hurt |
| Use diagonal covariance approximation instead of the full covariance matrix | NeuralUCB and NeuralTS · Shuttle | cumulative regret | No clear effect |
| Use perturbed rewards for exploration rather than diagonal-approximation confidence bonuses | Neural bandits · UCI classification datasets | cumulative regret | Helped |
| Use perturbed rewards for exploration rather than diagonal-approximation confidence bonuses | Neural bandits · LastFM | normalized cumulative reward | Helped |
| Use perturbed rewards for exploration rather than diagonal-approximation confidence bonuses | Neural bandits · Delicious | normalized cumulative reward | Mixed |
| Increase neural network width to 128, requiring diagonal covariance approximations for the baselines | NPR versus NeuralUCB and NeuralTS · Synthetic nonlinear reward functions | cumulative regret | Helped |
| Reveal all candidate arms rather than a subset each round | Neural bandit algorithms · Synthetic nonlinear reward functions | cumulative regret | Helped |
| Use scheduled lazy updates | FALCON+ · Synthetic nonlinear reward functions | cumulative regret | Hurt |
Closest papers by meaning
- Provably and Practically Efficient Neural Contextual Bandits (ICML 2023 · similarity 0.94)
- Exploration via Feature Perturbation in Contextual Bandits (NeurIPS 2025 · similarity 0.92)
- Contextual Bandits with Online Neural Regression (ICLR 2024 · similarity 0.90)
- Neural Dueling Bandits: Preference-Based Optimization with Human Feedback (ICLR 2025 · similarity 0.90)
- Robust Neural Contextual Bandit against Adversarial Corruptions (NeurIPS 2024 · similarity 0.90)
- Fast and Sample Efficient Multi-Task Representation Learning in Stochastic Contextual Bandits (ICML 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.