Research map / Optimization / Optimizers
Total stochastic gradient algorithms and applications in reinforcement learning
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 |
|---|---|---|---|
| Combine Gaussian shaping with BIW-LR and RP using total propagation | Model-based cart-pole policy search · Cart-pole swing-up, Tip Cost | success rate: 0.54 -> 0.6; 0.48 -> 0.69 | Helped |
| Increase model noise variance by 25 in GTP | GTP · Cart-pole swing-up, Angle Cost, k = 10^{-2} | success rate: 0.65 -> 0.88 | Helped |
| Increase model noise variance by 25 in GTP | GTP · Cart-pole swing-up, Tip Cost, k = 10^{-2} | success rate: 0.6 -> 0.8 | Helped |
| Use GTP rather than PILCO | Model-based cart-pole policy search · Cart-pole swing-up, Tip Cost | success rate: 0.92 -> 0.6; 0.73 -> 0.69 | Mixed |
| Use GTP rather than TP | Model-based cart-pole policy search · Cart-pole swing-up, Angle Cost | success rate: 0.82 -> 0.65; 0.99 -> 0.9 | Mixed |
| Use GTP rather than TP | Model-based cart-pole policy search · Cart-pole swing-up, Tip Cost | success rate: 0.54 -> 0.6; 0.48 -> 0.69 | Helped |
| Use Gaussian shaping with BIW-LR only, rather than combining BIW-LR and RP (ablation) | Gaussian shaping gradient · Cart-pole swing-up | success rate: Angle Cost: 0.42 vs 0.65 and 0.93 vs 0.9; Tip Cost: 0.45 vs 0.6 and 0.35 vs 0.69 | Mixed |
Closest papers by meaning
- A Baseline for Any Order Gradient Estimation in Stochastic Computation Graphs (ICML 2019 · similarity 0.91)
- Model-based Reinforcement Learning with Scalable Composite Policy Gradient Estimators (ICML 2023 · similarity 0.89)
- Statistically Efficient Off-Policy Policy Gradients (ICML 2020 · similarity 0.89)
- Learning Optimal Deterministic Policies with Stochastic Policy Gradients (ICML 2024 · similarity 0.88)
- Convergence Analysis of Policy Gradient Methods with Dynamic Stochasticity (ICML 2025 · similarity 0.88)
- On the Second-Order Convergence of Biased Policy Gradient Algorithms (ICML 2024 · similarity 0.88)
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.