AI Paper Atlas

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

GEX: A flexible method for approximating influence via Geometric Ensemble

SungYub Kim, Kyungsu Kim, Eunho Yang

NeurIPS 2023 · Learning theory and algorithms › Prediction and decision losses · approach: Active learning and subset selection · 1 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
remove linearization (use nonlinear Geometric Ensemble influence) (ablation)IGEX · CIFAR-10 noisy label detectionAUC: 64.11 ± 0.34 -> 99.74 ± 0.02Improved
remove linearization (ablation)IGEX · CIFAR-100 noisy label detectionAUC: 76.06 ± 0.36 -> 99.33 ± 0.03Improved
remove linearization (ablation)IGEX · CIFAR-10-N relabelingrelabeled accuracy: 91.04 ± 0.16 -> 93.54 ± 0.05Improved
reduce ensemble size M_GE from 32 to 8 (ablation)IGEX · CIFAR-10 noisy label detectionAP: 98.31 ± 0.06 -> 97.23 ± 0.11Hurt
reduce ensemble size M_GE from 32 to 4 (ablation)IGEX · CIFAR-100 noisy label detectionAP: 96.08 ± 0.12 -> 92.99 ± 0.18Hurt
IGEX instead of CLIGEX · CIFAR-10 noisy label detectionAUC: 57.60 ± 0.30 -> 99.74 ± 0.02Improved
IGEX instead of Deep-KNNIGEX · CIFAR-100-N noisy label detectionAUC: 71.59 ± 0.21 -> 89.76 ± 0.01Improved
use IArnoldi instead of random projection IRandProjIArnoldi · CIFAR-10 noisy label detectionAUC: 62.70 ± 0.19 -> 61.64 ± 0.13No clear effect
use checkpoint-based ITracInRP instead of IRandProjITracInRP · CIFAR-10 noisy label detectionAUC: 62.70 ± 0.19 -> 89.56 ± 0.14Improved
replace ITracInRP intermediate checkpoints with GE checkpoints (ablation)ITracInRP · CIFAR-100 noisy label detectionAUC: 74.99 ± 0.25 -> 88.84 ± 0.22Improved
replace ITracInRP intermediate checkpoints with GE checkpoints (ablation)ITracInRP · SVHN noisy label detectionAUC: 91.52 ± 0.16 -> 85.35 ± 0.20Hurt
relabel with Otsu threshold on CL self-influence in noisy real-world labelsCL · CIFAR-10-N relabelingrelabeled accuracy: 68.63 ± 0.32 -> 30.52 ± 0.02Hurt

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