Research map / Reinforcement learning / Imitation learning
Efficient Probabilistic Performance Bounds for Inverse Reinforcement Learning
Daniel Brown, Scott Niekum
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 posterior-sampled VaR bounds instead of worst-case feature-count bounds | Policy evaluation from demonstrations · Grid navigation | bound error: often four times tighter | Improved |
| Use the 0.95-VaR EVD bound instead of the Hoeffding-style EVD bound | Projection algorithm · Grid navigation | average EVD bound: 0.9372 vs 0.9372; 23,146 demonstrations vs 1 demonstration | Improved |
| Increase BIRL confidence parameter c on noisy demonstrations (ablation) | 0.95-VaR bound · Noisy grid navigation | bound error | Mixed |
| Set BIRL confidence parameter c to 50 for noisy demonstrations (ablation) | 0.95-VaR bound · Noisy grid navigation | accuracy | Hurt |
| Use BIRL confidence parameter c in the range (1,10] for noisy demonstrations (ablation) | 0.95-VaR bound · Noisy grid navigation | accuracy | Improved |
| Increase VaR quantile from 0.95 to 0.99 (ablation) | EVD confidence bound · Grid navigation | accuracy: 0.98 -> 1.0 | Improved |
| Increase VaR quantile from 0.95 to 0.99 (ablation) | EVD confidence bound · Grid navigation, 1 demonstration | average EVD bound: 0.9372 -> 1.1428 | Hurt |
| Use the VaR bound rather than WFCB to rank driving policies | Policy selection · Driving | policy ranking: right-safe: 3 -> 1; on-road: 1 -> 2; nasty: 2 -> 3 | Improved |
| Optimize the 0.99-VaR bound by changing policy actions one state at a time | BIRL maximum-likelihood policy · Two-feature navigation | 0.99-VaR | Improved |
| Use EVD instead of feature-count discrepancy to evaluate policies | Risk-aware policy bound · Partial or noisy demonstrations | bound applicability | Improved |
Closest papers by meaning
- Efficient Performance Bounds for Primal-Dual Reinforcement Learning from Demonstrations (ICML 2021 · similarity 0.90)
- Is Inverse Reinforcement Learning Harder than Standard Reinforcement Learning? A Theoretical Perspective (ICML 2024 · similarity 0.90)
- Recent Advancements in Inverse Reinforcement Learning (AAAI 2024 · similarity 0.89)
- Simplifying Constraint Inference with Inverse Reinforcement Learning (NeurIPS 2024 · similarity 0.88)
- Provably Efficient Exploration in Inverse Constrained Reinforcement Learning (ICML 2025 · similarity 0.88)
- Bayesian Exploration Networks (ICML 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.
