Research map / Reinforcement learning / Visual and representation RL
Unsupervised Learning of Goal Spaces for Intrinsically Motivated Goal Exploration
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 |
|---|---|---|---|
| learn the goal space from passive observations using AE, VAE, Isomap, or PCA | IMGEP-UGL · ArmBall and ArmArrow | KL-Coverage | Helped |
| use Radial Flow VAE for goal-space learning | IMGEP-UGL · ArmBall and ArmArrow | KL-Coverage | Hurt |
| use VAE-learned goals instead of engineered features | ArmBall · ArmBall | policies that move the ball: around 1600 -> around 2400 | Helped |
| increase the target embedding dimension from the true dimension to 10 (ablation) | RGE-Isomap · ArmBall | KL-Coverage | Hurt |
| increase the target embedding dimension from the true dimension to 10 (ablation) | RGE-AE · ArmBall | KL-Coverage | No clear effect |
| increase the target embedding dimension from the true dimension to 10 (ablation) | RGE-PCA · ArmBall | KL-Coverage | Helped |
| increase the target embedding dimension from the true dimension to 10 (ablation) | RGE-VAE · ArmBall | KL-Coverage | Helped |
| increase the target embedding dimension from the true dimension to 10 (ablation) | IMGEP-UGL algorithms · ArmArrow | KL-Coverage | Mixed |
| sample VAE goals from the isotropic Gaussian prior instead of KDE (ablation) | RGE-VAE · ArmBall and ArmArrow | KL-Coverage | No clear effect |
Closest papers by meaning
- Visual Reinforcement Learning with Imagined Goals (NeurIPS 2018 · similarity 0.88)
- Hierarchical Goal Abstractions via Learned Subset Relations (ICML 2026 · similarity 0.88)
- Exploring the Edges of Latent State Clusters for Goal-Conditioned Reinforcement Learning (NeurIPS 2024 · similarity 0.87)
- GoalLadder: Incremental Goal Discovery with Vision-Language Models (NeurIPS 2025 · similarity 0.86)
- PcLast: Discovering Plannable Continuous Latent States (ICML 2024 · similarity 0.85)
- Learning with a Mole: Transferable latent spatial representations for navigation without reconstruction (ICLR 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.