AI Paper Atlas

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

Selection by Prediction with Conformal p-values

Ying Jin, Emmanuel J. Candes

JMLR 2023 · Learning theory and algorithms › Prediction and decision losses · approach: Conformal prediction sets · 19 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 the clipped score V=M*1{y>0}-mu(x) instead of the residual score y-mu(x) (ablation)cfBH · HIV screening, q=0.1power: 0.0766 -> 0.0788Improved
use all calibration data (cfBH, BH_clip) instead of only Y=0 calibration samples (BH_sub) (ablation)cfBH · HIV screening, q=0.5power: 0.397 -> 0.410Improved
use the clipped score instead of the residual score on a 18-config drug-target interaction task (ablation)cfBH · DAVIS drug-target interactionpowerImproved
use all calibration data instead of Y=0 subset on the recruitment dataset (ablation)cfBH · Recruitment (Kaggle, 215 samples)powerImproved
apply Bonferroni correction instead of BH on the clipped score (ablation)cfBH · Simulations (8 settings, q=0.1)powerHurt
select by naive conformal prediction set equal to {1} instead of cfBHconformal prediction · HIV screeningFDR: about 80% at 1-alpha=0.90Hurt
increase noise level in simulated datacfBH · Simulations (settings 1-8)powerHurt
raise the population quantile threshold q_pop from 0.7 to 0.9cfBH · DAVIS drug-target interactionpowerHurt

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