AI Paper Atlas

Research map / Learning theory and algorithms / Prediction and decision losses

IALE: Imitating Active Learner Ensembles

Christoffer Löffler, Christopher Mutschler

NeurIPS 2022 · Learning theory and algorithms › Prediction and decision losses · approach: Active learning and subset selection

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 IALE instead of Random SamplingIALE · FMNISTaccuracy: 1-3% betterImproved
Use IALE instead of Random SamplingIALE · KMNISTaccuracy: 7-9% aheadImproved
Remove uncertainty experts (ablation)IALE · KMNISTaccuracyHurt
Remove soft-max uncertainty experts (ablation)IALE · KMNISTaccuracyHurt
Use fixed DAgger probability instead of exponential decay (ablation)IALE policy training · transfer to another datasetaccuracy: by a small marginImproved
Increase acquisition size while holding sub-pool size unchanged (ablation)IALE · MNISTaccuracyHurt
Increase sub-pool size to 1,000 (ablation)IALE · MNISTaccuracyImproved
Increase sub-pool size beyond its suitable upper limit (ablation)IALE · MNISTaccuracyHurt
Remove gradient representations from the state (ablation)IALE · active learning datasetsaccuracyHurt
Remove predicted-label representations from the state (ablation)IALE · active learning datasetsaccuracyNo clear effect
Use the full set of experts instead of leave-one-out expert sets (ablation)IALE · active learning datasetsaccuracyImproved

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