AI Paper Atlas

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

Adaptation Augmented Model-based Policy Optimization

Jian Shen, Hang Lai, Minghuan Liu, Han Zhao, Yong Yu, Weinan Zhang

JMLR 2023 · Reinforcement learning › Policy gradient and actor-critic · approach: Model-based RL with guarantees

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
add feature-based model adaptation (FAMPO) to MBPO (ablation)MBPO · Hopper, Walker2d, Ant (few interactions)average returnImproved
add instance-based model adaptation (IAMPO) to MBPO (ablation)MBPO · Hopper, Walker2d, Ant (few interactions)average returnImproved
add uncertainty-weighted sampling to model adaptation (ablation)FAMPO and IAMPO · Hopper and Walker2daverage returnImproved
add uncertainty-weighted sampling to model adaptation (ablation)FAMPO and IAMPO · Antaverage returnNo clear effect
add model adaptation to MBPOMBPO · six MuJoCo taskssample efficiency and returnImproved
add feature alignment loss (FAMPO) to MBPOMBPO · HopperWasserstein-1 distance of real vs simulated featuresImproved
add model adaptation to MBPOMBPO · compounding error vs trajectory length hcompounding errorImproved
increase adaptation iterations G2 too far (ablation)FAMPO · Hopper (epoch 40)average returnMixed
increase sampling temperature sigma too far (ablation)IAMPO · Walker2d (epoch 100)average returnHurt

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