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
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 changed | Where | Measured | Effect |
|---|---|---|---|
| instance-wise feature grouping instead of instance-wise selection without groups | L2X · Synthetic D1+D2 R1 | accuracy (%): 88 ± 4 -> 98.7 ± 1.3 | Improved |
| instance-wise grouping instead of global group learning | OSCAR · Synthetic D1+D2 R1 | accuracy (%): 50.0 ± 0.0 -> 98.7 ± 1.3 | Improved |
| gI vs L2X on 10-class MNIST | L2X · MNIST-10 | accuracy (%): 80.5 ± 2.5 -> 91.6 ± 0.85 | Improved |
| gI vs L2X on 2-class MNIST | L2X · MNIST-2 | accuracy (%): 97.1 ± 0.5 -> 96.7 ± 0.2 | Hurt |
| gI vs SHAP on MNIST-2 | SHAP · MNIST-2 | accuracy (%): 99.24 ± 0.46 -> 96.7 ± 0.2 | Hurt |
| gI vs Lasso on F-MNIST | Lasso · F-MNIST | accuracy (%): 96.7 ± 0.0 -> 94.6 ± 0.6 | Hurt |
| gI vs SHAP on D1 synthetic R1 | SHAP · Synthetic D1+D2 R1 | accuracy (%): 89.2 ± 0.97 -> 98.7 ± 1.3 | Improved |
| gI vs SHAP on D1+D2 R2 | SHAP · Synthetic D1+D2 R2 | accuracy (%): 94.4 ± 0.48 -> 100 ± 0 | Improved |
| increase number of groups k beyond the theoretical lower bound (ablation) | gI · Synthetic data | classification accuracy | No clear effect |
| use gI with m<k selected groups instead of all groups (ablation) | gI · Synthetic D1-D3 x R1-R3 | class accuracy: R1 D1 96.9 ± 1.5 vs 91.0 ± 3 | Mixed |
Closest papers by meaning
- A Rigorous Information-Theoretic Definition of Redundancy and Relevancy in Feature Selection Based on (Partial) Information Decomposition (JMLR 2023 · similarity 0.85)
- Composite Feature Selection Using Deep Ensembles (NeurIPS 2022 · similarity 0.85)
- Prior-Constrained Association Learning for Fine-Grained Generalized Category Discovery (AAAI 2025 · similarity 0.83)
- Unsupervised Concept Discovery Mitigates Spurious Correlations (ICML 2024 · similarity 0.81)
- Estimating Conditional Mutual Information for Dynamic Feature Selection (ICLR 2024 · similarity 0.81)
- Deep Taxonomic Networks for Unsupervised Hierarchical Prototype Discovery (NeurIPS 2025 · similarity 0.81)
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.
