Research map / Label-efficient and robust learning / Domain adaptation and generalization
Domain Adaptive Imitation Learning with Visual Observation
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 |
|---|---|---|---|
| Add image and feature reconstruction consistency to basic adversarial learning (ablation) | D3IL · RE3-to-two | average episodic return: -10.7 -> -6.0 | Helped |
| Add image-level cycle-consistency (ablation) | D3IL · RE3-to-two | average episodic return: -6.0 -> -3.0 | Helped |
| Add feature similarity (ablation) | D3IL · RE3-to-two | average episodic return: -3.0 -> -2.8 | Helped |
| Add feature-level cycle-consistency (ablation) | D3IL · RE3-to-two | average episodic return: -2.8 -> -2.6 | Helped |
| Replace external D_rew with BD_B (ablation) | D3IL · IP-to-color | average episodic return: 1000 ± 0 -> 293 ± 324 | Hurt |
| Replace external D_rew with BD_B (ablation) | D3IL · RE2-to-tilted | average episodic return: -4.3 ± 0.3 -> -11.9 ± 0.8 | Hurt |
| Use both image-level and feature-level cycle-consistency rather than either alone (ablation) | D3IL · RE2-to-three | behavior-feature domain independence | Helped |
| Remove feature-level cycle-consistency (ablation) | D3IL · RE2-to-three | behavior-feature domain independence | Hurt |
| Remove image-level cycle-consistency (ablation) | D3IL · RE2-to-three | desired behavior-feature extraction | Hurt |
| Remove domain encoders and their corresponding discriminators (ablation) | D3IL · IP-to-two, RE2-to-three, RE3-to-tilted, CartPoleBalance-to-Pendulum | average episodic return | Hurt |
Closest papers by meaning
- Cross-domain Imitation from Observations (ICML 2021 · similarity 0.91)
- Learn what matters: cross-domain imitation learning with task-relevant embeddings (NeurIPS 2022 · similarity 0.89)
- Diffusion Imitation from Observation (NeurIPS 2024 · similarity 0.87)
- Robust Visual Imitation Learning with Inverse Dynamics Representations (AAAI 2024 · similarity 0.87)
- Imitation Learning from Observation with Automatic Discount Scheduling (ICLR 2024 · similarity 0.87)
- OLLIE: Imitation Learning from Offline Pretraining to Online Finetuning (ICML 2024 · similarity 0.87)
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.