AI Paper Atlas

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

A Re-Balancing Strategy for Class-Imbalanced Classification Based on Instance Difficulty

Sihao Yu, Jiafeng Guo, Ruqing Zhang, Yixing Fan, Zizhen Wang, Xueqi Cheng

CVPR 2022 · Label-efficient and robust learning › Recognition and long-tail · approach: Group-robust classification · 46 citations (OpenAlex)

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
Use instance-level sampling weights based on learning-speed difficultyResNet-32 · Long-tailed CIFAR-10, imbalance ratio 100accuracy (%): 72.2 -> 75.0Improved
Use instance-level sampling weights based on learning-speed difficultyResNet-32 · Long-tailed CIFAR-100, imbalance ratio 100accuracy (%): 40.6 -> 42.3Improved
Combine TDE with instance-level re-samplingResNet-32 · Long-tailed CIFAR-100, imbalance ratio 100accuracy (%): 43.5 -> 44.9Improved
Increase the minority-class weights furtherClass-Balance Loss · Long-tailed CIFAR-100, imbalance ratio 100overall accuracy (%): 41.1 -> 38.2Hurt
Use instance-level difficulty re-samplingMultilayer Perceptron · Long-tailed MNIST binary classificationoverall accuracy (%): 86.04 -> 87.99Improved
Apply instance-level re-samplingLogistic Regression · UCI classification datasets (average)accuracy (%): 74.9 -> 78.2Improved
Apply instance-level re-samplingResNet-50 · iNaturalist 2019accuracy (%): 70.19% -> 71.08%Improved
Apply instance-level re-samplingResNet-101 · iNaturalist 2019accuracy (%): 72.84% -> 73.32%Improved

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