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

Self-supervised learning (SSL) as an effective paradigm of representationlearning has achieved tremendous success on various curated datasets in diversescenarios. Nevertheless, when facing the long-tailed distribution in real-worldapplications, it is still hard for existing methods to capture transferable androbust representation. Conventional SSL methods, pursuing sample-leveluniformity, easily leads to representation learning disparity where headclasses dominate the feature regime but tail classes passively collapse. Toaddress this problem, we propose a novel Geometric Harmonization (GH) method toencourage category-level uniformity in representation learning, which is morebenign to the minority and almost does not hurt the majority under long-taileddistribution. Specially, GH measures the population statistics of the embeddingspace on top of self-supervised learning, and then infer an fine-grainedinstance-wise calibration to constrain the space expansion of head classes andavoid the passive collapse of tail classes. Our proposal does not alter thesetting of SSL and can be easily integrated into existing methods in a low-costmanner. Extensive results on a range of benchmark datasets show theeffectiveness of GH with high tolerance to the distribution skewness. Our codeis available at https://github.com/MediaBrain-SJTU/Geometric-Harmonization.

Combating Representation Learning Disparity with Geometric Harmonization