AI Paper Atlas

Research map / Reinforcement learning / Imitation learning

Policy Learning Using Weak Supervision

Jingkang Wang, Hongyi Guo, Zhaowei Zhu, Yang Liu

NeurIPS 2021 · Reinforcement learning › Imitation learning · approach: Imitation combined with RL · 6 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
use peer reward instead of noisy rewardDDQN · CartPole e=0.1N_epi: 114.6 -> 86.2Improved
use peer reward instead of noisy rewardDDQN · CartPole e=0.4N_epi: 163.1 -> 106.9Improved
use peer reward instead of noisy rewardDDQN · CartPole e=0.3R_avg: 177.1 -> 174.1Hurt
use peer reward instead of surrogate rewardDDQN · CartPole e=0.3R_avg: 192.2 -> 174.1Hurt
use peer reward instead of true rewardDDQN · CartPole e=0.1N_epi: 101.2 -> 86.2Improved
use peer reward instead of noisy rewardDDQN · CartPole e=0.2R_avg: 168.8 -> 195.5Improved
use PeerBC (xi = 0.2) instead of standard BCBC · Atari Pongscore: 14.7 -> 18.8Improved
use PeerBC (xi = 0.5) instead of standard BCBC · Atari Enduroscore: 138.9 -> 230.9Improved
use PeerBC (xi = 0.5) instead of standard BCBC · Atari Boxingscore: 56.2 -> 75.6Improved
increase peer penalty xi from 0.5 to 1.0 (ablation)PeerBC · Atari Freewayscore: 22.4 -> 8.9Hurt
increase peer penalty xi from 0.2 to 0.5 (ablation)PeerBC · Atari lift over expertlift: +11.3% -> +19.5%Improved
use PeerBC instead of standard BCBC · CartPole-v1 imitation learningreturn: 18%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