AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

Instance-wise Feature Grouping

Aria Masoomi, Chieh Wu, Tingting Zhao, Zifeng Wang, Peter Castaldi, Jennifer Dy

NeurIPS 2020 · Label-efficient and robust learning › Recognition and long-tail · approach: Contrastive representation theory · 10 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
instance-wise feature grouping instead of instance-wise selection without groupsL2X · Synthetic D1+D2 R1accuracy (%): 88 ± 4 -> 98.7 ± 1.3Improved
instance-wise grouping instead of global group learningOSCAR · Synthetic D1+D2 R1accuracy (%): 50.0 ± 0.0 -> 98.7 ± 1.3Improved
gI vs L2X on 10-class MNISTL2X · MNIST-10accuracy (%): 80.5 ± 2.5 -> 91.6 ± 0.85Improved
gI vs L2X on 2-class MNISTL2X · MNIST-2accuracy (%): 97.1 ± 0.5 -> 96.7 ± 0.2Hurt
gI vs SHAP on MNIST-2SHAP · MNIST-2accuracy (%): 99.24 ± 0.46 -> 96.7 ± 0.2Hurt
gI vs Lasso on F-MNISTLasso · F-MNISTaccuracy (%): 96.7 ± 0.0 -> 94.6 ± 0.6Hurt
gI vs SHAP on D1 synthetic R1SHAP · Synthetic D1+D2 R1accuracy (%): 89.2 ± 0.97 -> 98.7 ± 1.3Improved
gI vs SHAP on D1+D2 R2SHAP · Synthetic D1+D2 R2accuracy (%): 94.4 ± 0.48 -> 100 ± 0Improved
increase number of groups k beyond the theoretical lower bound (ablation)gI · Synthetic dataclassification accuracyNo clear effect
use gI with m<k selected groups instead of all groups (ablation)gI · Synthetic D1-D3 x R1-R3class accuracy: R1 D1 96.9 ± 1.5 vs 91.0 ± 3Mixed

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