Research map / Reinforcement learning / Visual and representation RL
Near-Optimal Representation Learning for Hierarchical Reinforcement Learning
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 |
|---|---|---|---|
| use the proposed sub-optimality-bound representation objective instead of the evaluated baselines | goal-conditioned hierarchical RL · MuJoCo continuous-control navigation tasks | task performance | Helped |
| learn representations from raw top-down images with the proposed objective | hierarchical RL · Ant Maze Images | representation quality | Helped |
| use the proposed objective rather than VAE or E2C representation learning | hierarchical RL · MuJoCo navigation tasks | representation quality | Helped |
| use the proposed objective rather than end-to-end representation learning | hierarchical RL · MuJoCo continuous-control tasks | task performance | Helped |
| use the learned Ant Maze representation on a different environment | the proposed representation · Reflected Ant Maze and Ant Push | task performance | Helped |
| replace the distance-based energy with a dot-product energy (ablation) | the proposed representation objective · MuJoCo continuous-control tasks | task performance | Hurt |
| use other distance-based variants of the objective (ablation) | the proposed representation objective · MuJoCo continuous-control tasks | task performance | No clear effect |
| vary the beta-VAE hyperparameter (ablation) | VAE representation learning · MuJoCo tasks outside the Point environment | task performance | No clear effect |
| use a VAE representation instead of the proposed representation | hierarchical RL · high-information state-observation tasks | task performance | Hurt |
Closest papers by meaning
- Reconciling Spatial and Temporal Abstractions for Goal Representation (ICLR 2024 · similarity 0.90)
- Probabilistic Subgoal Representations for Hierarchical Reinforcement Learning (ICML 2024 · similarity 0.89)
- Learning Goal-Conditioned Representations for Language Reward Models (NeurIPS 2024 · similarity 0.88)
- Improving Unsupervised Hierarchical Representation with Reinforcement Learning (CVPR 2024 · similarity 0.88)
- Rethinking Model-based, Policy-based, and Value-based Reinforcement Learning via the Lens of Representation Complexity (NeurIPS 2024 · similarity 0.88)
- Deep Reinforcement Learning from Hierarchical Preference Design (ICML 2025 · similarity 0.88)
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.