Research map / Reinforcement learning / Policy gradient and actor-critic
Matryoshka Policy Gradient for Entropy-Regularized RL: Convergence and Global Optimality
François G. Ged, Maria Han Veiga
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 |
|---|---|---|---|
| use MPG instead of PG | REINFORCE · Frozen Lake (N=20) | cumulative reward | No clear effect |
| use MPG instead of softPG or nsPG | policy gradient · Frozen Lake (N=20) | cumulative reward | Improved |
| use MPG with shorter horizon N=10 or N=15 instead of PG, softPG, nsPG | policy gradient · Frozen Lake (N=10, 15) | treasure found | Improved |
| use MPG instead of PG/nsPG | policy gradient · Cart Pole (N=100) | test cumulative reward | No clear effect |
| add entropy regularisation (MPG, softPG) vs PG (ablation) | policy gradient · Cart Pole (N=100) | episodes to reach same reward | Hurt |
| use MPG instead of softPG | policy gradient · Cart Pole (N=100) | spread of test reward | Improved |
| use incomplete basis (4 of 5 features) in the analytical task (ablation) | MPG · Analytical task (5 states, n=2) | L-infinity error to optimal projected policy | No clear effect |
Closest papers by meaning
- Global linear convergence of entropy-regularized softmax policy gradient beyond tabular MDPs (NeurIPS 2026 · similarity 0.90)
- S$2$AC: Energy-Based Reinforcement Learning with Stein Soft Actor Critic (ICLR 2024 · similarity 0.88)
- Convergence Analysis of Policy Gradient Methods with Dynamic Stochasticity (ICML 2025 · similarity 0.88)
- Convergence Theorems for Entropy-Regularized and Distributional Reinforcement Learning (NeurIPS 2025 · similarity 0.88)
- Natural Policy Gradient as Doubly Smoothed Policy Iteration: A Bellman-Operator Framework (NeurIPS 2026 · similarity 0.88)
- DIME: Diffusion-Based Maximum Entropy Reinforcement Learning (ICML 2025 · 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.
