AI Paper Atlas

Research map / Reinforcement learning / Offline RL

Efficient Diffusion Policies For Offline Reinforcement Learning

Bingyi Kang, Xiao Ma, Chao Du, Tianyu Pang, Shuicheng YAN

NeurIPS 2023 · Reinforcement learning › Offline RL · approach: Simple practical offline RL · 12 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
Replace the Gaussian policy with EDP in TD3TD3 on Gym-locomotion D4RL · Gym-locomotion D4RL averageaverage normalized score (RAT): 75.3 -> 85.5Improved
Replace the Gaussian policy with EDP in IQLIQL on Kitchen D4RL · Kitchen D4RL averageaverage normalized score (RAT): 53.3 -> 59.4Improved
Replace the Gaussian policy with EDP in IQLIQL on Adroit D4RL · Adroit D4RL averageaverage normalized score (RAT): 54.4 -> 71.3Improved
Replace the Gaussian policy with EDP in IQLIQL on AntMaze D4RL · AntMaze D4RL averageaverage normalized score (RAT): 63.0 -> 73.4Improved
Use action approximation instead of full DDPM-chain training (ablation)EDP policy training · D4RL Gym-locomotiontraining speed: 2.3x trainingImproved
Use action approximation instead of full DDPM-chain training (ablation)EDP policy training · D4RL Gym-locomotionsampling speed: 3.3x samplingImproved
Replace DDPM sampling with DPM-Solver (ablation)EDP training · D4RL Gym-locomotiontraining speed: 2.3x training speedupImproved
Use action approximation at K=100 rather than full DDPM sampling (ablation)EDP on walker2d-medium-v2 · walker2d-medium-v2normalized score (OMS): 86.9 -> 85.5Hurt
Use action approximation at K=100 rather than full DDPM sampling (ablation)EDP on walker2d-medium-reply-v2 · walker2d-medium-reply-v2normalized score (OMS): 96.3 -> 93.3Hurt
Use action approximation at K=100 rather than full DDPM sampling (ablation)EDP on walker2d-medium-expert-v2 · walker2d-medium-expert-v2normalized score (OMS): 111.5 -> 111.1No clear effect
Increase the number of EAS candidate actions (ablation)EDP on Gym-locomotion · 8 of 9 locomotion tasksnormalized scoreImproved
Use EAS for evaluation with a normal TD3+BC policy (ablation)TD3+BC · Gym-locomotion D4RL averagenormalized scoreNo 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