AI Paper Atlas

Research map / Reinforcement learning / Policy gradient and actor-critic

Time Discretization-Invariant Safe Action Repetition for Policy Gradient Methods

Seohong Park, Jaekyeom Kim, Gunhee Kim

NeurIPS 2021 · Reinforcement learning › Policy gradient and actor-critic · approach: Variance-reduced policy gradients · 7 citations (OpenAlex)

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 changedWhereMeasuredEffect
Repeat actions until the state leaves a learned safe region instead of using a preset action durationPolicy-gradient methods on MuJoCo environments · Deterministic MuJoCo environments across varying δtotal rewardImproved
Use state-based safe regions rather than continuous action durationsSAR-PPO versus FiGAR-C-PPO · Stochastic MuJoCo environmentstotal rewardImproved
Use safe-region action repetition rather than per-step decisionsSAR-PPO versus DAU · Deterministic MuJoCo environments across varying δtotal rewardImproved
Use safe-region action repetitionSAR-PPO · Lowest-δ deterministic MuJoCo settingsdecision steps: 17.6× fewer decision steps on averageImproved
Use safe-region action repetition rather than standard PPOPPO · Deterministic MuJoCo environments at low δtotal rewardImproved
Scale PPO's learning rate by δ / δ₀PPO · Deterministic MuJoCo environments at low δtotal rewardNo clear effect
Use SAR rather than autoregressive action noisePPO · Deterministic MuJoCo environments at low δtotal rewardImproved
Use safe-region action repetition rather than FiGAR-CA2C · Deterministic MuJoCo environmentstotal rewardImproved
Use safe-region action repetition rather than FiGAR-CTRPO · Stochastic MuJoCo environmentstotal rewardImproved
Add temporal thresholds to safe-region criteria (ablation)SAR-PPO · Four stochastic POMDP MuJoCo environmentstotal rewardImproved
Remove the upper limit on action duration (ablation)SAR · MuJoCo environmentstotal rewardMixed
Decrease δ while using SARSAR · Swimmer-v2total rewardHurt

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy