AI Paper Atlas

Research map / Reinforcement learning / Offline RL

Data-Efficient Pipeline for Offline Reinforcement Learning with Limited Data

Allen Nie, Yannis Flet-Berliac, Deon Richmond Jordan, William Steenbergen, Emma Brunskill

NeurIPS 2022 · Reinforcement learning › Offline RL · approach: Simple practical offline RL · 2 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
Use SSR-RRS-5 instead of CV-5 for algorithm-hyperparameter selectionOffline RL policy selection · Robomimic Can-Pairedaverage true return: 0.72 -> 0.73Improved
Use SSR-RRS-5 instead of SSR-RRS-2 for algorithm-hyperparameter selectionOffline RL policy selection · Robomimic Transportaverage true return: 0.62 -> 0.70Improved
Use SSR-RRS-5 instead of a single-split baselineOffline RL policy selection · D4RL Hopper medium-expertaverage true return: 3481.34 -> 3657.80Improved
Use SSR-RRS-5 instead of CV-2 for algorithm-hyperparameter selectionOffline RL policy selection · Sepsis-MDP (n=200)average true return: -10.26 -> -7.85Improved
Use SSR-RRS-5 instead of SSR-RRS-2 for algorithm-hyperparameter selectionOffline RL policy selection · Sepsis-POMDP (n=1000)average true return: 2.40 -> 6.75Improved
Increase the number of repeated random splits from 1 to 5 (ablation)SSR-RRS · Sepsis-POMDPselected policy's true performanceImproved
Retrain the selected algorithm-hyperparameter pair on the full dataset instead of only half (ablation)SSR-selected policy · Small-to-medium offline RL datasetstrue rewardImproved
Use repeated random subsampling rather than one split (ablation)Offline RL algorithm-hyperparameter selection · Sepsis-POMDP (n=5000)selected policy performanceImproved
Use CV-5 rather than SSR-RRS-5Offline RL policy selection · D4RL Hopper randomaverage true return: 325.37 -> 325.37No clear effect

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