Research map / Reinforcement learning / Imitation learning
Strictly Batch Imitation Learning by Energy-based Distribution Matching
Daniel Jarrett, Ioana Bica, Mihaela van der Schaar
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 energy-based occupancy loss to behavioral cloning | EDM · Acrobot-v1, CartPole-v1, LunarLander-v2, BeamRider-v4 (1 demo) | average return: Acrobot-v1: 0.796 -> 0.896; CartPole-v1: 0.321 -> 0.428; LunarLander-v2: 0.575 -> 0.633; BeamRider-v4: 0.124 -> 0.486 | Improved |
| Add the energy-based occupancy loss to behavioral cloning | EDM · MIMIC-III-2a | ACC: 0.861 -> 0.891 | Improved |
| Add the energy-based occupancy loss to behavioral cloning | EDM · MIMIC-III-4a | ACC: 0.696 -> 0.720 | Improved |
| Use EDM rather than the strongest baseline at 15 demos | EDM · CartPole-v1 | average return: 0.998 ± 0.001 vs. 0.998 ± 0.002 | No clear effect |
| Use EDM rather than the strongest baseline at 15 demos | EDM · LunarLander-v2 | average return: 0.970 ± 0.028 vs. 0.970 ± 0.033 | No clear effect |
| Use EDM rather than the strongest baseline | EDM · MIMIC-III-2a and MIMIC-III-4a | AUC: MIMIC-III-2a: 0.915 -> 0.922; MIMIC-III-4a: 0.864 -> 0.873 | Improved |
| Use an off-policy adaptation that predicts next-state feature maps | DSFN · Gym environments | average return | Hurt |
Closest papers by meaning
- Noise-conditioned Energy-based Annealed Rewards (NEAR): A Generative Framework for Imitation Learning from Observation (ICLR 2025 · similarity 0.86)
- Markov Balance Satisfaction Improves Performance in Strictly Batch Offline Imitation Learning (AAAI 2025 · similarity 0.86)
- Causal Imitation for Markov Decision Processes: a Partial Identification Approach (NeurIPS 2024 · similarity 0.85)
- Agnostic Interactive Imitation Learning: New Theory and Practical Algorithms (ICML 2024 · similarity 0.85)
- Learning When to Stop: Selective Imitation Learning Under Arbitrary Dynamics Shift (NeurIPS 2026 · similarity 0.85)
- Imitation Learning from Purified Demonstrations (ICML 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.
