Research map / Reinforcement learning / Visual and representation RL
Meta-Reinforcement Learning with Self-Modifying Networks
Mathieu Chalvidal, Thomas Serre, Rufin VanRullen
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 |
|---|---|---|---|
| MetODS versus RL2 on 6x6 maze | MetODS · Maze 6x6 | accumulated reward: 149.3 -> 169.1 | Improved |
| MetODS versus RL2 on 8x8 maze | MetODS · Maze 8x8 | accumulated reward: 72.1 -> 87.3 | Improved |
| MetODS versus RL2 on 10x10 unseen maze | MetODS · Maze 10x10 | accumulated reward: 28.1 -> 34.9 | Improved |
| MetODS versus RL2 on 12x12 unseen maze | MetODS · Maze 12x12 | accumulated reward: 11.1 -> 13.9 | Improved |
| MetODS versus MAML on 8x8 maze | MetODS · Maze 8x8 | accumulated reward: 14.9 -> 87.3 | Improved |
| add element-wise tuning of plasticity to MAML (ablation) | MAML · Maze exploration | accumulated reward | Hurt |
| add element-wise plasticity alpha to MetODS (ablation) | MetODS · Maze exploration | accumulated reward | Improved |
| increase recursivity depth S of the update rule (ablation) | MetODS · Maze exploration | accumulated reward | Improved |
| transform rightmost vector of writing equation with a linear projection (ablation) | MetODS · Maze exploration | accumulated reward | Improved |
| transform rightmost vector of writing equation with an MLP (ablation) | MetODS · Maze exploration | accumulated reward | No clear effect |
| increase network size to N=20 neurons | MetODS · Harlow task | reward | Improved |
| MetODS versus MAML and PEARL on Ant-dir adaptation within a single episode | MetODS · MuJoCo Ant-dir | reward | Improved |
Closest papers by meaning
- Meta-Learning through Hebbian Plasticity in Random Networks (NeurIPS 2020 · similarity 0.89)
- A contrastive rule for meta-learning (NeurIPS 2022 · similarity 0.89)
- Transformers are Meta-Reinforcement Learners (ICML 2022 · similarity 0.87)
- Meta-Dynamical State Space Models for Integrative Neural Data Analysis (ICLR 2025 · similarity 0.86)
- MetaCARD: Meta-Reinforcement Learning with Task Uncertainty Feedback via Decoupled Context-Aware Reward and Dynamics Components (AAAI 2024 · similarity 0.85)
- Decoupling Meta-Reinforcement Learning with Gaussian Task Contexts and Skills (AAAI 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.
