Research map / Reinforcement learning / Policy gradient and actor-critic
Genetic-Gated Networks for Deep Reinforcement Learning
Simyung Chang, John Yang, Jaeseok Choi, Nojun Kwak
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 genetic gating (G2AC) to A2C | A2C · Atari AMIDAR | average return: 548.2 -> 929.0 | Improved |
| add genetic gating (G2AC) to A2C | A2C · Atari ASSAULT | average return: 2026.6 -> 15,147.2 | Improved |
| add genetic gating (G2AC) to A2C | A2C · Atari SEAQUEST | average return: 1763.7 -> 1,772.6 | No clear effect |
| add genetic gating (G2AC) to A2C | A2C · Atari VENTURE | average return: 0.0 -> 0.0 | No clear effect |
| G2AC at 200M frames vs A2C at 320M frames | A2C · 50 Atari games | games won: better in 39 games, same in three, worse in eight | Improved |
| replace genetic algorithm with random binary gates (ablation) | G2AC/G2PPO · Atari and MuJoCo | learning curve | Hurt |
| separate GA optimization from network training (ablation) | G2AC/G2PPO · Atari and MuJoCo | learning speed | Hurt |
| add genetic gating to PPO8 (G2PPO) | PPO8 · MuJoCo | final score | Improved |
| G2PPO early training vs PPO8 | PPO8 · MuJoCo | early learning progress | Hurt |
| add genetic gating to A2C: wall-clock cost | A2C · five Atari games | steps per second: 3668 -> 3572 | Hurt |
| add genetic gating to PPO8: wall-clock cost | PPO8 · MuJoCo | steps per second: 1143 -> 1110 | Hurt |
| raise mutation probability from 0.03 to 0.1 and 0.3 (ablation) | G2PPO · MuJoCo | performance: 0.03 -> 0.1 and 0.3 | Hurt |
Closest papers by meaning
- Symbolic Regression via Deep Reinforcement Learning Enhanced Genetic Programming Seeding (NeurIPS 2021 · similarity 0.86)
- Efficient Reinforcement Learning by Discovering Neural Pathways (NeurIPS 2024 · similarity 0.85)
- Policy Optimization by Genetic Distillation (ICLR 2018 · similarity 0.85)
- GFlowNet Training by Policy Gradients (ICML 2024 · similarity 0.84)
- Neural Genetic Search in Discrete Spaces (ICML 2025 · similarity 0.84)
- Gradient Boosting Reinforcement Learning (ICML 2025 · similarity 0.83)
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.
