Research map / Reinforcement learning / Policy gradient and actor-critic
A Geometric Perspective on Optimal Representations for Reinforcement Learning
Marc Bellemare, Will Dabney, Robert Dadashi, Adrien Ali Taiga, Pablo Samuel Castro, Nicolas Le Roux, Dale Schuurmans, Tor Lattimore, Clare Lyle
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 adversarial value functions (AVFs) as auxiliary tasks instead of the value of one policy (ablation) | deep representation learning, d=16 · four-room domain | learned representation quality (qualitative) | Improved |
| use value functions of random deterministic policies as auxiliary tasks instead of AVFs (ablation) | deep representation learning, d=16 · four-room domain | learned representation quality (qualitative) | Hurt |
| use AVF representation instead of single value-function representation (ablation) | model-based SARSA · four-room domain | average discounted return | Improved |
| use AVF representation versus proto-value functions | model-based SARSA · four-room domain | average discounted return | No clear effect |
| increase the number of auxiliary tasks (ablation) | AVF representation · four-room domain | average discounted return | Improved |
Closest papers by meaning
- On Value Function Representation of Long Horizon Problems (AAAI 2018 · similarity 0.89)
- Value Function Spaces: Skill-Centric State Abstractions for Long-Horizon Reasoning (ICLR 2022 · similarity 0.88)
- Near-optimal Offline Reinforcement Learning with Linear Representation: Leveraging Variance Information with Pessimism (ICLR 2022 · similarity 0.88)
- Linear Feature Encoding for Reinforcement Learning (NeurIPS 2016 · similarity 0.88)
- The Value Function Polytope in Reinforcement Learning (ICML 2019 · similarity 0.88)
- Stable Offline Value Function Learning with Bisimulation-based Representations (ICML 2025 · similarity 0.86)
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.
