AI Paper Atlas

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

Open Category Detection with PAC Guarantees

Si Liu, Risheek Garrepalli, Thomas Dietterich, Alan Fern, Dan Hendrycks

ICML 2018 · Learning theory and algorithms › Prediction and decision losses · approach: Adaptive testing and experiments · 46 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
Increase the mixture sample size nOpen Category Detection with Isolation Forest · Synthetic datarecall: 0.710±0.033 -> 0.918±0.010 (α=0.01, n=100 -> n=10000)Improved
Increase the alien fraction α in the mixture training dataOpen Category Detection with Isolation Forest · Synthetic data (n=10000, target recall 95%)recall: 0.918±0.010 -> 0.950 ±0.001 (α=0.01 -> α=0.50)Improved
Use an overestimate α′ instead of the true alien fraction αOpen Category Detection · UCI datasetsfalse positive rate: from 0.01 to 0.336Hurt
Use an overestimate α′ instead of the true alien fraction αOpen Category Detection · UCI datasetsrecall: from 0.01 to 0.05Improved
Add isotonization and clipping to the estimated alien CDF (ablation)Open Category Detection · Experiments reported in supplementary materialsoverall performanceNo clear effect
Increase the alien fraction α in the mixture training dataOpen Category Detection with Isolation Forest · UCI and vision benchmark datasetsfalse positive rateImproved
Use the method on small datasets or when α is smallOpen Category Detection · Synthetic and benchmark datasetsrecallHurt

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