Research map / Reinforcement learning / Policy gradient and actor-critic
Online Nonstochastic Model-Free 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 |
|---|---|---|---|
| Use action-space exploration rather than policy-space exploration | Bandit GPC · Bandit linear control, convex costs | regret upper bound: O(\sqrt{d_x d_u d_{min}} T^{3/4}) -> O(\sqrt{d_u d_{min}} T^{3/4}) | Helped |
| Use action-space exploration rather than policy-space exploration | Bandit GPC · Bandit linear control, smooth costs | regret upper bound: The paper says the improvement carries over to smooth costs but gives no paired prior bound here. | Helped |
| Add pseudo-disturbance policies using PD2 or PD3 | DDPG · Noisy Hopper, Noisy Walker 2D, and Noisy Ant | episode return | Helped |
| Add the PD1 pseudo-disturbance estimate | DDPG · Noisy Hopper and other RL settings | episode return | Mixed |
| Add MF-GPC with PD1 | DDPG · Noisy Hopper | episode return | Helped |
| Use RBPC rather than BPC | Linear dynamical systems · High-dimensional LDS | episode return | Helped |
| Use RBPC rather than BPC | Linear dynamical systems · Low-dimensional LDS | episode return | No clear effect |
Closest papers by meaning
- Online Learning in Stabilized Linear Dynamical Games with Adversarial Disturbances (NeurIPS 2026 · similarity 0.91)
- Online Control with Adversarial Disturbance for Continuous-time Linear Systems (NeurIPS 2024 · similarity 0.90)
- Online Control with Adversarial Disturbances (ICML 2019 · similarity 0.89)
- Adversarial Diffusion for Robust Reinforcement Learning (NeurIPS 2025 · similarity 0.89)
- Enhancing Robustness in Deep Reinforcement Learning: A Lyapunov Exponent Approach (NeurIPS 2024 · similarity 0.88)
- Beyond Worst-case Attacks: Robust RL with Adaptive Defense via Non-dominated Policies (ICLR 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.