Research map / Reinforcement learning / Imitation learning
Revisiting Distribution Correction Estimation for Offline Imitation Learning with Suboptimal Dataset
Quang Anh Pham, Tien Anh Mai, Akshat Kumar
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 |
|---|---|---|---|
| add the mixture distribution (vs removing it) (ablation) | ReDICE · D4RL offline IL tasks | average return: removing it causes the largest drop (about 20 times) | Improved |
| add the Gumbel regression loss (vs removing it) (ablation) | ReDICE · D4RL offline IL tasks | average return | Improved |
| add tau-scaled policy extraction (vs removing it) (ablation) | ReDICE · D4RL offline IL tasks | average return | Improved |
| raise tau up to 0.9 (ablation) | ReDICE · D4RL offline IL tasks | average return: peaks at tau = 0.9 | Improved |
| raise tau beyond 0.9 toward 1.0 (ablation) | ReDICE · D4RL offline IL tasks | average return: diminishing or negative returns after 0.9 | Hurt |
| set mixture coefficient beta to 0.5 rather than 0.1 or 0.9 (ablation) | ReDICE · D4RL offline IL tasks | average return: beta = 0.5 best vs 0.1 and 0.9 | Improved |
| larger beta (0.7 or 0.9) on low-coverage data (ablation) | ReDICE · random+few-expert vs medium+few-expert | average return | Mixed |
| add tau-scaled policy extraction | ReCOIL · D4RL offline IL tasks | average return: 67.76 -> 81.49 | Improved |
| replace ReCOIL objective (with tau-scaling) by the ReDICE objective | ReCOIL with tau-scaling · D4RL offline IL tasks | average return: 81.49 -> 97.68 | Improved |
| use ReDICE instead of ReCOIL | offline IL, 1 expert trajectory · D4RL suboptimal datasets average | normalized score: 67.76 -> 97.68 | Improved |
| use ReDICE instead of ReCOIL under limited expert coverage | offline IL, no expert rollouts · D4RL limited coverage average | normalized score: 28.16 -> 38.45 | Improved |
| use ReDICE instead of baselines on maritime navigation | ShipNaviSim · maritime navigation | goal reach rate: ≈60% -> ≈85% | Improved |
Closest papers by meaning
- Mitigating Covariate Shift in Behavioral Cloning via Robust Stationary Distribution Correction (NeurIPS 2024 · similarity 0.91)
- ODICE: Revealing the Mystery of Distribution Correction Estimation via Orthogonal-gradient Update (ICLR 2024 · similarity 0.91)
- GO-DICE: Goal-Conditioned Option-Aware Offline Imitation Learning via Stationary Distribution Correction Estimation (AAAI 2024 · similarity 0.90)
- Offline Imitation from Observation via Primal Wasserstein State Occupancy Matching (ICML 2024 · similarity 0.90)
- OLLIE: Imitation Learning from Offline Pretraining to Online Finetuning (ICML 2024 · similarity 0.89)
- How to Leverage Diverse Demonstrations in Offline Imitation Learning (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.
