AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

How Should an Agent Practice?

Janarthanan Rajendran, Richard Lewis, Vivek Veeriah, Honglak Lee, Satinder Singh

AAAI 2020 · Reinforcement learning › Visual and representation RL · approach: Reward shaping for sparse rewards · 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
Add practice periods with meta-gradient-learned practice rewardsA2C · Pongmean rewardImproved
Add practice periods with meta-gradient-learned practice rewardsA2C · PacManmean rewardImproved
Replace meta-gradient reward learning with a fixed random practice reward (ablation)A2C + Meta-Gradients Practice · Pongmean rewardHurt
Replace meta-gradient reward learning with a fixed random practice reward (ablation)A2C + Meta-Gradients Practice · PacManmean rewardHurt
Replace meta-gradient-learned rewards with match-reward predictionA2C + Meta-Gradients Practice · Pongmean rewardHurt
Replace meta-gradient-learned rewards with match-reward predictionA2C + Meta-Gradients Practice · PacManmean rewardMixed
Use practice time for meta-gradient-guided practice instead of additional matchesA2C · Pongmean rewardImproved
Use practice time for practice without ghosts instead of additional matchesA2C · PacManmean rewardHurt
Add match-reward training after practice updates to adapt the intrinsic practice rewardREINFORCE · Corridor worldreward per episodeImproved

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