AI Paper Atlas

Research map / Reinforcement learning / Offline RL

Supported Trust Region Optimization for Offline Reinforcement Learning

Yixiu Mao, Hongchang Zhang, Chen Chen, Yi Xu, Xiangyang Ji

ICML 2023 · Reinforcement learning › Offline RL · approach: Offline RL sample complexity theory · 1 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 a Gaussian behavior density estimator with a conditional VAE (ablation)STR · D4RL tasks, especially hopper-randomnormalized scoreMixed
Initialize the policy randomly instead of from the pretrained behavior policy (ablation)STR · D4RL tasks, especially hopper-expert and hopper-medium-expertnormalized scoreMixed
Use vanilla importance sampling instead of self-normalized importance sampling (ablation)STR · D4RL tasksnormalized scoreHurt
Use a support constraint rather than a density constraint toward behaviorSTR versus prior EAWBC methods · halfcheetah-randomnormalized scoreImproved
Use supported trust-region updates with importance sampling from the datasetSTR versus offline baselines · AntMaze tasksnormalized scoreImproved
Use supported trust-region updates rather than AWAC updatesSTR versus AWAC · hopper-medium, hopper-medium-replay, walker2d-medium-replaynormalized scoreMixed
Use STR policy iterations with approximate Q and sampling errorSTR · Offline policy updatessafe policy improvementImproved

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