Research map / Label-efficient and robust learning / Recognition and long-tail
Neural Collapse for Unconstrained Feature Model under Cross-entropy Loss with Imbalanced Data
Wanli Hong, Shuyang Ling
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 |
|---|---|---|---|
| increase within-class sample size nA with fixed imbalance ratio | ResNet18 mean prediction vectors · Cifar10, kA=kB=5, r=2 | mean pairwise correlation: converge toward -1/9 as nA grows from 500 to 5500 | Improved |
| use theoretical minority-collapse threshold lambda_Z = sqrt(nB)/N | UFM theory · ResNet18 Cifar10, kA=kB=5 | minority collapse threshold: theory 1/300 vs empirical 0.0033 | Improved |
| use theoretical complete-collapse threshold lambda_Z = sqrt(nA)/N | UFM theory · ResNet18 Cifar10, kA=kB=5 | complete collapse threshold: theory 0.075 vs empirical complete collapse at some lambda_Z <= 0.069 | Hurt |
| use theoretical complete-collapse threshold | UFM theory · ResNet18 Cifar10, kA=3, kB=7 | complete collapse threshold: theory 0.0102 vs empirical between 0.0086 and 0.0094 | Hurt |
| use theoretical minority-collapse threshold | UFM theory · ResNet18 Cifar10, kA=3, kB=7 | minority collapse threshold: predicted 0.0045 matches empirical | Improved |
| vary nA from 100 to 1400 at lambda_Z=0.005 | theory thresholds in nA · ResNet18 Cifar10, kA=kB=5 | collapse thresholds in nA: minority 300 confirmed; complete 1392 overestimated | Mixed |
| more imbalanced dataset (Dataset_2 vs Dataset_1) | trained network vs UFM solution · VGG11/VGG13 mean prediction matrix | relative error: 0.04 -> approximately 0.1 | Hurt |
Closest papers by meaning
- Neural Collapse for Cross-entropy Class-Imbalanced Learning with Unconstrained ReLU Features Model (ICML 2024 · similarity 0.94)
- Neural Collapse To Multiple Centers For Imbalanced Data (NeurIPS 2024 · similarity 0.90)
- Neural Collapse in Cumulative Link Models for Ordinal Regression: An Analysis with Unconstrained Feature Model (NeurIPS 2025 · similarity 0.90)
- Wide Neural Networks Trained with Weight Decay Provably Exhibit Neural Collapse (ICLR 2025 · similarity 0.89)
- Neural Collapse Beyond the Unconstrained Features Model: Landscape, Dynamics, and Generalization in the Mean-Field Regime (ICML 2025 · similarity 0.89)
- Understanding Representation of Deep Equilibrium Models from Neural Collapse Perspective (NeurIPS 2024 · similarity 0.89)
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.
