AI Paper Atlas

Research map / Reinforcement learning / Offline RL

Weighted model estimation for offline model-based reinforcement learning

Toru Hishinuma, Kei Senda

NeurIPS 2021 · Reinforcement learning › Offline RL · approach: Offline RL sample complexity theory · 3 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
weighted model estimation (Algorithm 1) instead of unweighted ERMpendulum model · Pendulum policy evaluationestimated discounted return (true -15.7): -80.7 -> -17.8Improved
add weight α = 0.2 (ablation)Algorithm 3 · Walker2d-medium-expertnormalized score: 59.5 ± 49.4 -> 108.2 ± 0.5Improved
add weight α = 0.2 (ablation)Algorithm 3 · Walker2d-mediumnormalized score: 81.7 ± 1.2 -> 60.7 ± 29.0Hurt
add weight α = 0.2 (ablation)Algorithm 3 · HalfCheetah-medium-expertnormalized score: 73.9 ± 24.2 -> 85.7 ± 21.6Improved
add weight α = 0.2 (ablation)Algorithm 3 · Hopper-medium-replaynormalized score: 97.2 ± 10.9 -> 104.0 ± 3.2Improved
add weight α = 0.2 (ablation)Algorithm 3 · Hopper-medium-expertnormalized score: 109.3 ± 1.1 -> 104.9 ± 10.1No clear effect
add weight α = 0.2 (ablation)Algorithm 3 · HalfCheetah-medium-replaynormalized score: 72.1 ± 1.4 -> 65.5 ± 6.4Hurt
add weight α = 0.2 (ablation)Algorithm 3 · HalfCheetah-mediumnormalized score: 75.7 ± 1.5 -> 73.1 ± 5.2No clear effect
add weight α = 0.2 (ablation)Algorithm 3 · Walker2d-medium-replaynormalized score: 80.7 ± 3.1 -> 82.7 ± 3.3No clear effect
add weight α = 0.2 (ablation)Algorithm 3 · Hopper-randomnormalized score: 30.2 ± 4.4 -> 32.7 ± 0.5Improved
simplified weighted loss (Algorithm 2) versus ERM start of iterationAlgorithm 2 · Pendulumweighted lossImproved
weighted model estimation (α = 0 MOPO-like) versus original MOPOAlgorithm 3 with α = 0 · Hopper-mediumnormalized score: 28.0 ± 12.4 -> 100.9 ± 2.7Improved

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