AI Paper Atlas

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

Imbalance Trouble: Revisiting Neural-Collapse Geometry

Christos Thrampoulidis, Ganesh Ramachandra Kini, Vala Vakilian, Tina Behnia

NeurIPS 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Loss design and classifier theory · 13 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
measure convergence to SELI geometry instead of ETF geometrydeep nets (ResNet-18, VGG-13) · STEP-imbalanced MNIST, Fashion-MNIST, CIFAR10distance to geometry (classifiers, embeddings, logits)Improved
increase imbalance ratio R (up to 100)deep nets · STEP-imbalanced MNIST, Fashion-MNIST, CIFAR10convergence rate to SELIHurt
measure SELI vs ETF for SGD on the unconstrained feature modelUFM trained with SGD · synthetic UFM, k=4distance to geometryImproved
vanishing ridge regularization in the UFM (ablation)UFM global minimizers · synthetic UFM, k=4distance to SELIImproved
increase imbalance ratio R in UFM regularization path (ablation)UFM global minimizers · synthetic UFM, k=4rate of convergence to SELI as lambda vanishesHurt
classifier versus embedding geometrydeep nets and UFM · STEP-imbalanced datasetsconvergence to SELIMixed

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