Research map / Bandits and online learning / Contextual bandits and Thompson sampling
Efficient Counterfactual Learning from Bandit Feedback
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 an estimated propensity score instead of the known true score | Self-normalized inverse probability weighting for ad CTR evaluation · CyberAgent ads — existing logging and best-action policies | Confidence interval width: 6-34% | Helped |
| Use the best advertisement predicted for each context instead of the logging policy | CyberAgent ad policy · CyberAgent ads | CTR: 10-15% | Helped |
| Estimate propensity scores with ridge logistic regression rather than use the true score | Self-normalized inverse probability weighting · CyberAgent ads — existing logging policy | CI length reduction: 0 -> -30.2% | Helped |
| Estimate propensity scores with ridge logistic regression rather than use the true score | Self-normalized inverse probability weighting · CyberAgent ads — choose best action by context | CI length reduction: 0 -> -34.1% | Helped |
| Use gradient boosting rather than random forest to estimate the propensity score | Self-normalized inverse probability weighting · CyberAgent ads — existing logging policy | CI length reduction: -6.62% -> -22.5% | Helped |
Closest papers by meaning
- Counterfactual Learning with General Data-Generating Policies (AAAI 2023 · similarity 0.92)
- Bayesian Counterfactual Risk Minimization (ICML 2019 · similarity 0.91)
- Cost-Effective Incentive Allocation via Structured Counterfactual Inference (AAAI 2020 · similarity 0.91)
- Log-Sum-Exponential Estimator for Off-Policy Evaluation and Learning (ICML 2025 · similarity 0.89)
- Logarithmic Smoothing for Pessimistic Off-Policy Evaluation, Selection and Learning (NeurIPS 2024 · similarity 0.89)
- Robustly Improving Bandit Algorithms with Confounded and Selection Biased Offline Data: A Causal Approach (AAAI 2024 · similarity 0.88)
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.