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

Data heterogeneity is an inherent challenge that hinders the performance offederated learning (FL). Recent studies have identified the biased classifiersof local models as the key bottleneck. Previous attempts have used classifiercalibration after FL training, but this approach falls short in improving thepoor feature representations caused by training-time classifier biases.Resolving the classifier bias dilemma in FL requires a full understanding ofthe mechanisms behind the classifier. Recent advances in neural collapse haveshown that the classifiers and feature prototypes under perfect trainingscenarios collapse into an optimal structure called simplex equiangular tightframe (ETF). Building on this neural collapse insight, we propose a solution tothe FL's classifier bias problem by utilizing a synthetic and fixed ETFclassifier during training. The optimal classifier structure enables allclients to learn unified and optimal feature representations even underextremely heterogeneous data. We devise several effective modules to betteradapt the ETF structure in FL, achieving both high generalization andpersonalization. Extensive experiments demonstrate that our method achievesstate-of-the-art performances on CIFAR-10, CIFAR-100, and Tiny-ImageNet.

No Fear of Classifier Biases: Neural Collapse Inspired Federated Learning with Synthetic and Fixed Classifier