Research map / Probabilistic and scientific modeling / Molecules and proteins
Learning Tree Interpretation from Object Representation for Deep Reinforcement Learning
Guiliang Liu, Xiangyu Sun, Oliver Schulte, Pascal Poupart
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 IMONet latent features instead of VAE latent features (ablation) | MCRTS · Flappy Bird | VR: 7.83E-2 -> 8.53E-2 | Improved |
| use IMONet latent features instead of VAE latent features (ablation) | MCRTS · Space Invaders | VR: 4.82E-2 -> 5.37E-2 | Improved |
| use IMONet latent features instead of VAE latent features (ablation) | MCRTS · Assault | VR: 6.58E-2 -> 7.53E-2 | Improved |
| replace CART with MCRTS split search on IMONet features (ablation) | IMONet tree · Flappy Bird | VR-PL: 4.02E-4 -> 1.37E-3 | Improved |
| replace CART with MCRTS split search on IMONet features (ablation) | IMONet tree · Assault | Number of leaves: 315 -> 83 | Improved |
| use IMONet features with MCRTS instead of raw-input CART (ablation) | tree mimic · Flappy Bird | Number of leaves: 1007 -> 62 | Improved |
| use IMONet features with MCRTS instead of raw-input CART (ablation) | tree mimic · Space Invaders | Variance Reduction (VR): 4.96E-2 -> 5.37E-2 | Mixed |
| use IMONet+MCRTS instead of raw-input M5-RT (ablation) | tree mimic · Flappy Bird | Variance Reduction (VR): 9.59E-2 -> 8.53E-2 | Hurt |
| use VAE features instead of raw input for CART (ablation) | CART · Flappy Bird | Variance Reduction (VR): 8.51E-2 -> 7.25E-2 | Hurt |
| use IMONet features instead of VAE for VR-LMT (ablation) | VR-LMT · Space Invaders | Variance Reduction (VR): 3.52E-2 -> 4.54E-2 | Improved |
| MCRTS with look-ahead split vs greedy linear model tree (ablation) | IMONet tree · Assault | VR-PL: 2.27E-4 -> 9.07E-4 | Improved |
| use latent mimic tree instead of super-pixel tree for human interpretability | IMONet+MCRTS · Human evaluation | preference (%): 83 | Improved |
Closest papers by meaning
- SkillTree: Explainable Skill-Based Deep Reinforcement Learning for Long-Horizon Control Tasks (AAAI 2025 · similarity 0.85)
- RGMDT: Return-Gap-Minimizing Decision Tree Extraction in Non-Euclidean Metric Space (NeurIPS 2024 · similarity 0.85)
- Enhancing Interpretability in Deep Reinforcement Learning through Semantic Clustering (NeurIPS 2025 · similarity 0.84)
- RICE: Breaking Through the Training Bottlenecks of Reinforcement Learning with Explanation (ICML 2024 · similarity 0.84)
- Improving Unsupervised Hierarchical Representation with Reinforcement Learning (CVPR 2024 · similarity 0.84)
- ORCAID: Oblique Rule-Based Continuous-Action Interpretation for Deep RL Policies (NeurIPS 2026 · similarity 0.84)
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.
