AI Paper Atlas

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

JMLR 2024 · Label-efficient and robust learning › Recognition and long-tail · approach: Loss design and classifier theory · 22 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
increase within-class sample size nA with fixed imbalance ratioResNet18 mean prediction vectors · Cifar10, kA=kB=5, r=2mean pairwise correlation: converge toward -1/9 as nA grows from 500 to 5500Improved
use theoretical minority-collapse threshold lambda_Z = sqrt(nB)/NUFM theory · ResNet18 Cifar10, kA=kB=5minority collapse threshold: theory 1/300 vs empirical 0.0033Improved
use theoretical complete-collapse threshold lambda_Z = sqrt(nA)/NUFM theory · ResNet18 Cifar10, kA=kB=5complete collapse threshold: theory 0.075 vs empirical complete collapse at some lambda_Z <= 0.069Hurt
use theoretical complete-collapse thresholdUFM theory · ResNet18 Cifar10, kA=3, kB=7complete collapse threshold: theory 0.0102 vs empirical between 0.0086 and 0.0094Hurt
use theoretical minority-collapse thresholdUFM theory · ResNet18 Cifar10, kA=3, kB=7minority collapse threshold: predicted 0.0045 matches empiricalImproved
vary nA from 100 to 1400 at lambda_Z=0.005theory thresholds in nA · ResNet18 Cifar10, kA=kB=5collapse thresholds in nA: minority 300 confirmed; complete 1392 overestimatedMixed
more imbalanced dataset (Dataset_2 vs Dataset_1)trained network vs UFM solution · VGG11/VGG13 mean prediction matrixrelative error: 0.04 -> approximately 0.1Hurt

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