AI Paper Atlas

Research map / Reinforcement learning / Imitation learning

Information Directed Reward Learning for Reinforcement Learning

David Lindner, Matteo Turchetta, Sebastian Tschiatschek, Kamil Ciosek, Andreas Krause

NeurIPS 2021 · Reinforcement learning › Imitation learning · approach: Imitation combined with RL · 0 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
update candidate policies 20 times instead of uniform samplingIDRL · Swimmer-Corridorregret after 20 queries: 11.8 -> 2.4Improved
replace IGR with IDRL (20 updates)IDRL · Ant-Corridorregret after 20 queries: 17.2 -> 2.2Improved
replace EIR with IDRL (20 updates)IDRL · Ant-Corridorregret after 20 queries: 17.5 -> 2.2Improved
reduce candidate policy updates from 20 to 4 (ablation)IDRL · Swimmer-Corridorregret after 20 queries: 2.4 -> 2.8Hurt
reduce candidate policy updates from 20 to 1 (ablation)IDRL · Ant-Corridorregret after 20 queries: 2.2 -> 12Hurt
reduce candidate policy updates from 20 to 1 (ablation)IDRL · Swimmer-Corridorruntime (hours): 40 -> 20Improved
reduce candidate policy updates from 20 to 1 (ablation)IDRL · Ant-Corridorruntime (hours): 40 -> 10Improved
single-update IDRL versus baselines (ablation)IDRL (1 update) · Swimmer-Corridorregret after 20 queries: 11.8 -> 7.6Improved
select queries by IDRL instead of pure information gain (IGR)IDRL · DriverregretImproved
select queries by IDRL instead of EPD and IGRIDRL · GridworldsregretImproved
IDRL instead of baselinesIDRL · Swimmer-Corridorcosine similarity of learned rewardNo clear effect
IDRL instead of uniform/IGRIDRL · MuJoCo Deep RL, 1400 comparisonsnormalized scoreImproved

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