English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Class imbalance is a common challenge in real-world recognition tasks, wherethe majority of classes have few samples, also known as tail classes. Weaddress this challenge with the perspective of generalization and empiricallyfind that the promising Sharpness-Aware Minimization (SAM) fails to addressgeneralization issues under the class-imbalanced setting. Through investigatingthis specific type of task, we identify that its generalization bottleneckprimarily lies in the severe overfitting for tail classes with limited trainingdata. To overcome this bottleneck, we leverage class priors to restrict thegeneralization scope of the class-agnostic SAM and propose a class-awaresmoothness optimization algorithm named Imbalanced-SAM (ImbSAM). With theguidance of class priors, our ImbSAM specifically improves generalizationtargeting tail classes. We also verify the efficacy of ImbSAM on twoprototypical applications of class-imbalanced recognition: long-tailedclassification and semi-supervised anomaly detection, where our ImbSAMdemonstrates remarkable performance improvements for tail classes and anomaly.Our code implementation is available athttps://github.com/cool-xuan/Imbalanced_SAM.

ImbSAM: A Closer Look at Sharpness-Aware Minimization in Class-Imbalanced Recognition