AI Paper Atlas

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

An Effective Meaningful Way to Evaluate Survival Models

Shi-ang Qi, Neeraj Kumar, Mahtab Farrokh, Weijie Sun, Li-Hao Kuan, Rajesh Ranganath, Ricardo Henao, Russell Greiner

ICML 2023 · Learning theory and algorithms › Prediction and decision losses · approach: Data attribution and recourse

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
replace hinge/uncensored/IPCW de-censoring with pseudo-observation (MAE-PO) surrogate event timesMAE evaluation of survival models · 29 semi-synthetic datasetsnumber of experiments where metric is best: 22 of 29 (76%)Improved
use margin-time de-censoring (MAE-margin) instead of pseudo-observationsMAE evaluation under exponential censoring · exponential censoring datasetsnumber of experiments where metric is best: best in 3 of 5 exponential datasets vs 2 for POMixed
MAE-hinge as the censored-subject estimatorMAE evaluation · semi-synthetic datasets with high censoringcloseness to true MAEHurt
pseudo-observation de-censoring vs uncensored-only evaluationMAE evaluation under feature-dependent censoring · GBM dataset feature-dependent censoringcloseness to true MAEMixed
use KM population mean survival time as surrogate event time (MAE-Pop-PO) instead of individual pseudo-observations (ablation)MAE evaluation · semi-synthetic datasetscloseness to true MAENo 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