Research map / Reinforcement learning / Policy gradient and actor-critic
Direct Advantage Estimation
Hsiao-Ru Pan, Nico Gürtler, Alexander Neitz, Bernhard Schölkopf
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 |
|---|---|---|---|
| replace GAE with DAE | PPO · MinAtar (5 envs, baseline network) | Overall/Last score: 6.6 -> 10.5 | Improved |
| replace GAE with DAE | PPO · ALE (baseline network) | environments won: 10 -> 32 | Improved |
| replace GAE with DAE, Wide network | PPO · ALE | environments won (Overall): 8 -> 35 | Improved |
| replace GAE with DAE, Deep network | PPO · ALE | environments won (Overall): 6 -> 35 | Improved |
| widen network layers | DAE · MinAtar Breakout | Last score: 844.7 -> 2409.5 | Improved |
| widen network layers | GAE · MinAtar Breakout | Last score: 168.2 -> 637.7 | Improved |
| widen network layers | DAE · MinAtar SpaceInvaders | Last score: 1675.3 -> 11743.3 | Improved |
| widen network layers | GAE · MinAtar SpaceInvaders | Last score: 504.2 -> 1058.0 | Improved |
| replace GAE with DAE | PPO · synthetic environment | MSE of advantage estimate | Improved |
| replace GAE with DAE | PPO · synthetic environment, early training | expected return | Hurt |
| replace GAE with DAE | PPO · synthetic environment, late training | expected return | Improved |
| replace DAE loss with Indirect (separate Q and V) (ablation) | DAE · MinAtar | normalized score | Hurt |
Closest papers by meaning
- Skill or Luck? Return Decomposition via Advantage Functions (ICLR 2024 · similarity 0.89)
- Adaptive Advantage-Guided Policy Regularization for Offline Reinforcement Learning (ICML 2024 · similarity 0.87)
- GAE Falls Short in Imperfect-Information Self-Play Reinforcement Learning (NeurIPS 2026 · similarity 0.86)
- A2PO: Towards Effective Offline Reinforcement Learning from an Advantage-aware Perspective (NeurIPS 2024 · similarity 0.86)
- Discerning Temporal Difference Learning (AAAI 2024 · similarity 0.85)
- Off-policy estimation with adaptively collected data: the power of online learning (NeurIPS 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.
