Research map / Reinforcement learning / Imitation learning
Towards Theoretical Understanding of Inverse Reinforcement Learning
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 uniform sampling with the stated confidence bounds | US-IRL for feasible reward set estimation · Generative model, time-inhomogeneous transitions | sample complexity: \widetilde{O} \left( \frac{H^3SA}{\epsilon^2}\left( \log\left( \frac{1}{\delta} \right) +S \right)\right) | Helped |
| Use uniform sampling with the stated confidence bounds | US-IRL for feasible reward set estimation · Generative model, time-homogeneous transitions | sample complexity: \widetilde{O} \left( \frac{H^2SA}{\epsilon^2}\left( \log\left( \frac{1}{\delta} \right) +S \right)\right) | Helped |
| Restrict rewards to be state-only | Feasible reward set estimation · Finite-horizon IRL with ε-close transition models | Hausdorff distance between feasible reward sets: \mathcal{H}_{d^{\text{G}}}(\mathcal{R}_{\text{state}} ,\widehat{\mathcal{R}}_{\text{state}} ) \ge 1 | Hurt |
| Restrict rewards to be time-homogeneous | Feasible reward set estimation · Finite-horizon IRL with ε-close transition models | Hausdorff distance between feasible reward sets: \mathcal{H}_{d^{\text{G}}}(\mathcal{R}_{\text{hom}} ,\widehat{\mathcal{R}}_{\text{hom}}) \ge 1/4 | Hurt |
| Require a β-margin reward | Feasible reward set estimation · Finite-horizon IRL with ε-close transition models | non-emptiness of the feasible reward set | Hurt |
| Allow any bounded reward function rather than imposing state-only, time-homogeneous, or β-margin restrictions | Feasible reward set estimation · Finite-horizon IRL | Lipschitz continuity of feasible reward sets: \mathcal{H}_{d^{\text{G}}}(\mathcal{R},\widehat{\mathcal{R}}) \le \frac{2 \rho^{\text{G}}((\mathcal{M},\pi^E),(\widehat{ | Helped |
Closest papers by meaning
- Recent Advancements in Inverse Reinforcement Learning (AAAI 2024 · similarity 0.96)
- Offline Inverse RL: New Solution Concepts and Provably Efficient Algorithms (ICML 2024 · similarity 0.94)
- Is Inverse Reinforcement Learning Harder than Standard Reinforcement Learning? A Theoretical Perspective (ICML 2024 · similarity 0.93)
- How does Inverse RL Scale to Large State Spaces? A Provably Efficient Approach (NeurIPS 2024 · similarity 0.93)
- Sub-optimal Experts mitigate Ambiguity in Inverse Reinforcement Learning (NeurIPS 2024 · similarity 0.92)
- Towards the Transferability of Rewards Recovered via Regularized Inverse Reinforcement Learning (NeurIPS 2024 · similarity 0.90)
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.