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

Learning with Noisy Labels (LNL) has become an appealing topic, asimperfectly annotated data are relatively cheaper to obtain. Recentstate-of-the-art approaches employ specific selection mechanisms to separateclean and noisy samples and then apply Semi-Supervised Learning (SSL)techniques for improved performance. However, the selection step mostlyprovides a medium-sized and decent-enough clean subset, which overlooks a richset of clean samples. To fulfill this, we propose a novel LNL framework ProMixthat attempts to maximize the utility of clean samples for boosted performance.Key to our method, we propose a matched high confidence selection techniquethat selects those examples with high confidence scores and matched predictionswith given labels to dynamically expand a base clean sample set. To overcomethe potential side effect of excessive clean set selection procedure, wefurther devise a novel SSL framework that is able to train balanced andunbiased classifiers on the separated clean and noisy samples. Extensiveexperiments demonstrate that ProMix significantly advances the currentstate-of-the-art results on multiple benchmarks with different types and levelsof noise. It achieves an average improvement of 2.48\% on the CIFAR-N dataset.The code is available at https://github.com/Justherozen/ProMix

ProMix: Combating Label Noise via Maximizing Clean Sample Utility