Research map / Reinforcement learning / Imitation learning
Learning from a Learner
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 |
|---|---|---|---|
| Recover reward from consecutive soft policy iterations (Theorem 2) and plan with it | Grid world observer · Grid world | score J(pi) | Helped |
| Observer learns from learner's trajectories via reward regression (LfL) instead of imitation | Grid world observer across learner algorithms · Grid world (SPI, SVI, Q-learning, random improvements) | score J(pi): observer beats learner for SPI, SVI, random; slightly below for Q-learning (3.65 vs 3.99) | Mixed |
| Use PPO-based learner trajectories and recovered reward for observer training | Continuous control observer · Reacher, HalfCheetah, Ant | normalized score | Helped |
| Recover reward from learner trajectories (LfL) on Hopper | Hopper observer · Hopper | normalized score | Hurt |
| Use state-only reward regression instead of state-action reward (ablation) | Grid world recovered reward · Grid world | score J(pi): 5.68 for both state-action and state-only rewards | No clear effect |
Closest papers by meaning
- Hybrid Inverse Reinforcement Learning (ICML 2024 · similarity 0.90)
- Is Inverse Reinforcement Learning Harder than Standard Reinforcement Learning? A Theoretical Perspective (ICML 2024 · similarity 0.90)
- Recent Advancements in Inverse Reinforcement Learning (AAAI 2024 · similarity 0.90)
- Environment Design for Inverse Reinforcement Learning (ICML 2024 · similarity 0.90)
- In-Trajectory Inverse Reinforcement Learning: Learn Incrementally Before an Ongoing Trajectory Terminates (NeurIPS 2024 · similarity 0.89)
- Offline Inverse RL: New Solution Concepts and Provably Efficient Algorithms (ICML 2024 · similarity 0.89)
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.