AI Paper Atlas

Research map / Reinforcement learning / Imitation learning

Learning Noise-Induced Reward Functions for Surpassing Demonstrations in Imitation Learning

Liangyu Huo, Zulin Wang, Mai Xu

AAAI 2023 · Reinforcement learning › Imitation learning · approach: Behavior cloning theory and models · 0 citations (OpenAlex)

Read the paper

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 weighted preference learning with a nonlinear weight based on noise-level differencesLERP · HalfCheetah, Hopper, Walker2d, Humanoid, Enduro, Kung Fu Master, Pong, Seaquest, Space Invaders, Ms Pacmanaverage returnImproved
Use weighted preference learning with a nonlinear weight based on noise-level differencesLERP · Antaverage return: 738.9 -> 612.7Hurt
Use weighted preference learning with a nonlinear weight based on noise-level differencesLERP · Q*bertaverage return: 27967.5 -> 9622.5Hurt
Replace the nonlinear noise weight with a linear weight (ablation)LERP · HalfCheetah unseen demonstrationsreturn correlation: 0.830 -> 0.899Improved
Replace the nonlinear noise weight with a linear weight (ablation)LERP · Hopper unseen demonstrationsreturn correlation: -0.830 -> -0.624Improved
Replace the nonlinear noise weight with a linear weight (ablation)LERP · Ant unseen demonstrationsreturn correlation: 0.913 -> 0.963Improved
Replace the nonlinear noise weight with a linear weight (ablation)LERP · Walker2d unseen demonstrationsreturn correlation: 0.421 -> -0.122Hurt
Replace the nonlinear noise weight with a linear weight (ablation)LERP · Humanoid unseen demonstrationsreturn correlation: 0.608 -> 0.964Improved
Use the nonlinear noise weight rather than the linear-weight variant (ablation)LERP reward model · MuJoCo preference pairspreference accuracy: 76.3% -> 82.0%Improved
Use the nonlinear noise weight rather than the linear-weight variant (ablation)LERP reward model · MuJoCo preference pairspreference accuracy: 89.9% -> 92.7%Improved

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