Research map / Reinforcement learning / Policy gradient and actor-critic
Policy Evaluation with Latent Confounders via Optimal Balance
Andrew Bennett, Nathan Kallus
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 optimal latent-proxy balancing weights instead of IPS weights based on observed X | Policy evaluation · Synthetic step, n=2000 | RMSE: $.40\pm.01$ -> $.08\pm.01$ | Improved |
| Use latent-proxy balancing rather than balancing observed X | Optimal balancing policy evaluation · Synthetic step, n=2000 | RMSE: $2.0\pm.01$ -> $.08\pm.01$ | Improved |
| Set regularization γ to 5.0 rather than 0.001 (ablation) | OptZ · Synthetic step, n=2000 | RMSE: $.08\pm.01$ -> $.17\pm.01$ | Hurt |
| Set regularization γ to 0.2 rather than 0.001 (ablation) | OptZ · Synthetic step, n=2000 | RMSE: $.08\pm.01$ -> $.08\pm.01$ | No clear effect |
| Set regularization γ to 1.0 rather than 0.001 (ablation) | OptZ · Synthetic step, n=2000 | RMSE: $.08\pm.01$ -> $.09\pm.01$ | Hurt |
| Fit the latent outcome function directly instead of optimizing balancing weights | Policy evaluation · Synthetic step, n=2000 | RMSE: $.08\pm.01$ -> $2.6\pm.01$ | Hurt |
Closest papers by meaning
- Confounding-Robust Policy Improvement (NeurIPS 2018 · similarity 0.88)
- Logarithmic Smoothing for Pessimistic Off-Policy Evaluation, Selection and Learning (NeurIPS 2024 · similarity 0.88)
- Two-way Deconfounder for Off-policy Evaluation in Causal Reinforcement Learning (NeurIPS 2024 · similarity 0.85)
- Policy Learning for Balancing Short-Term and Long-Term Rewards (ICML 2024 · similarity 0.85)
- Off-policy estimation with adaptively collected data: the power of online learning (NeurIPS 2024 · similarity 0.85)
- Mixed-Effects Contextual Bandits (AAAI 2024 · similarity 0.85)
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.
