Research map / Reinforcement learning / Policy gradient and actor-critic
Guided Policy Search via Approximate Mirror Descent
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 |
|---|---|---|---|
| Replace BADMM with MDGPS using off-policy sampling and classic step size | Guided policy search · Blind Peg Insertion, iteration 12 | success rate: 75.00~% -> 83.33~% | Helped |
| Replace BADMM with MDGPS using off-policy sampling and global step size | Guided policy search · Blind Peg Insertion, iteration 12 | success rate: 75.00~% -> 91.67~% | Helped |
| Use off-policy sampling instead of on-policy sampling with classic step size (ablation) | MDGPS · Blind Peg Insertion, iteration 12 | success rate: 58.33~% -> 83.33~% | Helped |
| Use off-policy sampling instead of on-policy sampling with global step size (ablation) | MDGPS · Blind Peg Insertion, iteration 12 | success rate: 58.33~% -> 91.67~% | Helped |
| Use the global step size instead of classic step size with off-policy sampling (ablation) | MDGPS · Blind Peg Insertion, iteration 12 | success rate: 83.33~% -> 91.67~% | Helped |
| Use the global step size instead of classic step size with on-policy sampling (ablation) | MDGPS · Blind Peg Insertion, iteration 12 | success rate: 58.33~% -> 58.33~% | No clear effect |
| Replace BADMM with MDGPS using off-policy sampling and classic step size | Guided policy search · Peg Insertion, iteration 12 | success rate: 77.78~% -> 70.73~% | Hurt |
| Replace BADMM with MDGPS using off-policy sampling and global step size | Guided policy search · Peg Insertion, iteration 12 | success rate: 77.78~% -> 92.59~% | Helped |
| Use the global step size instead of classic step size with off-policy sampling (ablation) | MDGPS · Peg Insertion, iteration 12 | success rate: 70.73~% -> 92.59~% | Helped |
| Use off-policy sampling instead of on-policy sampling with global step size (ablation) | MDGPS · Peg Insertion, iteration 12 | success rate: 85.19~% -> 92.59~% | Helped |
| Use the global step size instead of classic step size with on-policy sampling (ablation) | MDGPS · Peg Insertion, iteration 12 | success rate: 92.59~% -> 85.19~% | Hurt |
| Replace BADMM with MDGPS | Guided policy search · Obstacle Navigation and Peg Insertion | average distance to target or hole | No clear effect |
Closest papers by meaning
- Policy Mirror Descent with Lookahead (NeurIPS 2024 · similarity 0.89)
- Diffusion Policy Policy Optimization (ICLR 2025 · similarity 0.88)
- Learning mirror maps in policy mirror descent (ICLR 2025 · similarity 0.88)
- Flow Matching Policy Optimization with Mirror Descent and Entropy Constraints (NeurIPS 2026 · similarity 0.87)
- Convergence of Policy Mirror Descent Beyond Compatible Function Approximation (ICML 2025 · similarity 0.87)
- Operator World Models for Reinforcement Learning (NeurIPS 2024 · similarity 0.87)
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.