Research map / Reinforcement learning / MDPs and planning
Lifelong Learning with a Changing Action Set
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 |
|---|---|---|---|
| Reuse the cardinality-invariant policy component instead of reinitializing the policy after each action-set change | LAICA · Maze and recommender system domains | performance | Helped |
| Infer action-space structure during adaptation when new actions are added | LAICA · Maze and recommender system domains | performance | Helped |
| Use actor-critic to optimize the decision component rather than DPG (ablation) | LAICA · Maze and recommender system domains | performance | Helped |
| Add newly available actions to the existing policy without adapting an action-space structure model | Baseline(2) · Maze and recommender system domains | performance | Hurt |
| Use an action-cardinality-dependent policy that is reinitialized after each action-set change | Baseline(1) · Maze and recommender system domains | data efficiency | Hurt |
| Use the action-structure-based policy parameterization before any new actions are added | LAICA · Maze and recommender system domains | performance | Helped |
| Use Baseline(2) in the sparse-reward maze task | Baseline(2) · Maze | performance consistency | Hurt |
Closest papers by meaning
- Reset-Free Lifelong Learning with Skill-Space Planning (ICLR 2021 · similarity 0.87)
- Provably Efficient Lifelong Reinforcement Learning with Linear Representation (ICLR 2023 · similarity 0.87)
- Model-based Lifelong Reinforcement Learning with Bayesian Exploration (NeurIPS 2022 · similarity 0.86)
- Autonomous Option Invention for Continual Hierarchical Reinforcement Learning and Planning (AAAI 2025 · similarity 0.85)
- Continual Reinforcement Learning by Planning with Online World Models (ICML 2025 · similarity 0.85)
- Self-Composing Policies for Scalable Continual Reinforcement Learning (ICML 2024 · similarity 0.84)
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.