Research map / Learning theory and algorithms / Optimal transport
On the Benefit of Optimal Transport for Curriculum Reinforcement Learning
Pascal Klink, Carlo D’Eramo, Jan Peters, Joni Pajarinen
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 entropy lower bound on task distribution (H_LB 0.5) (ablation) | CURROT · E-Maze, pseudo-metric d_P* | final return: 0 -> 0.45 | Improved |
| add entropy lower bound (H_LB 0.5) (ablation) | CURROT · E-Maze, shortest-path distance | final return: 0.62 -> 0.61 | No clear effect |
| add entropy lower bound (H_LB 1.0) (ablation) | CURROT · E-Maze, shortest-path distance | final return: 0.62 -> 0.53 | Hurt |
| add entropy regularization (lambda 1e-2) (ablation) | GRADIENT · E-Maze, pseudo-metric d_P* | final return: 0.55 -> 0.30 | Hurt |
| add entropy regularization (lambda 1e-4) (ablation) | GRADIENT · E-Maze, shortest-path distance | final return: 0.60 -> 0.62 | No clear effect |
| use Euclidean distance instead of shortest-path distance (ablation) | CURROT and GRADIENT · E-Maze | final return: 0.62 -> 0 | Hurt |
| use pseudo-metric d_P* instead of shortest-path distance (ablation) | CURROT · E-Maze | final return: 0.62 -> 0 | Hurt |
Closest papers by meaning
- Curriculum Reinforcement Learning via Constrained Optimal Transport (ICML 2022 · similarity 0.99)
- Curriculum Reinforcement Learning using Optimal Transport via Gradual Domain Adaptation (NeurIPS 2022 · similarity 0.91)
- Curriculum Learning for Reinforcement Learning Domains: A Framework and Survey (JMLR 2020 · similarity 0.90)
- A Probabilistic Interpretation of Self-Paced Learning with Applications to Reinforcement Learning (JMLR 2021 · similarity 0.90)
- Diffusion-based Curriculum Reinforcement Learning (NeurIPS 2024 · similarity 0.88)
- PORTAL: Automatic Curricula Generation for Multiagent Reinforcement Learning (AAAI 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.
