Research map / Reinforcement learning / Visual and representation RL
ERL-Re$^2$: Efficient Evolutionary Reinforcement Learning with Shared State Representation and Individual Policy Representation
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 ERL-Re² with shared state and individual linear policy representations | EA and RL policy optimization · Six MuJoCo tasks | convergence rate vs. EA: 5x | Helped |
| Use behavior-level rather than parameter-level crossover and mutation (ablation) | ERL-Re²(TD3) · MuJoCo tasks | return | Helped |
| Use both behavior-level crossover and behavior-level mutation (ablation) | ERL-Re²(TD3) · MuJoCo tasks | return | Helped |
| Estimate fitness with PeVFA H-step bootstrapping instead of the RL critic and replay-buffer states (ablation) | ERL-Re²(TD3) · MuJoCo tasks | return | Helped |
| Add the PeVFA fitness surrogate (ablation) | ERL-Re² · Swimmer | return | Helped |
| Update the shared state representation using both the RL critic and PeVFA (ablation) | ERL-Re²(TD3) · MuJoCo tasks | return | Helped |
| Tune the number of sampled EA policies K (ablation) | ERL-Re²(TD3) · MuJoCo tasks | return | Mixed |
| Use one rollout episode for the fitness surrogate (ablation) | ERL-Re²(TD3) · MuJoCo tasks | return | No clear effect |
| Increase the population size from 5 to 10 (ablation) | ERL-Re²(TD3) · MuJoCo tasks | return | Hurt |
| Add shared state and individual policy representations to EA without the RL side (ablation) | EA · MuJoCo tasks | return | Helped |
| Add mutual-information self-supervised loss to the shared representation (ablation) | ERL-Re²(TD3) · Ant and Walker | return | Mixed |
| Optimize the shared representation using current and previous-generation policy representations (ablation) | ERL-Re² · MuJoCo tasks | return | No clear effect |
Closest papers by meaning
- ERL-TD: Evolutionary Reinforcement Learning Enhanced with Truncated Variance and Distillation Mutation (AAAI 2024 · similarity 0.92)
- EvoRainbow: Combining Improvements in Evolutionary Reinforcement Learning for Policy Search (ICML 2024 · similarity 0.89)
- Value-Evolutionary-Based Reinforcement Learning (ICML 2024 · similarity 0.89)
- Learning to Reuse Policies in State Evolvable Environments (ICML 2025 · similarity 0.89)
- Two-Stage Evolutionary Reinforcement Learning for Enhancing Exploration and Exploitation (AAAI 2024 · similarity 0.89)
- Evolutionary Reinforcement Learning with Parameterized Action Primitives for Diverse Manipulation Tasks (AAAI 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.