AI Paper Atlas

Research map / Reinforcement learning / Visual and representation RL

Learn What Not to Learn: Action Elimination with Deep Reinforcement Learning

Tom Zahavy, Matan Haroush, Nadav Merlis, Daniel J Mankowitz, Shie Mannor

NeurIPS 2018 · Reinforcement learning › Visual and representation RL · approach: Scalable RL systems and environments · 125 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 AE-DQN instead of prior workOpen Zork with A3 · Open Zork, A3cumulative reward: 13.5 -> 39Improved
Use AE-DQN instead of prior workOpen Zork with A4 · Open Zork, A4cumulative reward: 8.8 -> 16Improved
Train AE-DQN for 2M rather than 1M stepsOpen Zork with A3 · Open Zork, A3cumulative reward: 39 -> 44Improved
Train AE-DQN for 2M rather than 1M stepsOpen Zork with A4 · Open Zork, A4cumulative reward: 16 -> 16No clear effect
Add action elimination to Q-learningTabular Q-learning · Grid Worldepisode performanceImproved
Add action elimination to DQNDQN · Troll Questcumulative rewardImproved
Use action elimination with the larger action spaceAE-DQN versus DQN · Egg Quest, A2 with T=100learning speedImproved
Increase the number of state categories, increasing invalid actionsTabular action elimination Q-learning · Grid Worldlearning performanceImproved
Use an almost completely random elimination signalTabular action elimination Q-learning · Grid Worldlearning performanceNo clear effect

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