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 been extensively studied, with existingapproaches typically following a framework that alternates between clean sampleselection and semi-supervised learning (SSL). However, this approach has alimitation: the clean set selected by the Deep Neural Network (DNN) classifier,trained through self-training, inevitably contains noisy samples. This mixtureof clean and noisy samples leads to misguidance in DNN training during SSL,resulting in impaired generalization performance due to confirmation biascaused by error accumulation in sample selection. To address this issue, wepropose a method called Collaborative Sample Selection (CSS), which leveragesthe large-scale pre-trained model CLIP. CSS aims to remove the mixed noisysamples from the identified clean set. We achieve this by training a2-Dimensional Gaussian Mixture Model (2D-GMM) that combines the probabilitiesfrom CLIP with the predictions from the DNN classifier. To further enhance theadaptation of CLIP to LNL, we introduce a co-training mechanism with acontrastive loss in semi-supervised learning. This allows us to jointly trainthe prompt of CLIP and the DNN classifier, resulting in improved featurerepresentation, boosted classification performance of DNNs, and reciprocalbenefits to our Collaborative Sample Selection. By incorporating auxiliaryinformation from CLIP and utilizing prompt fine-tuning, we effectivelyeliminate noisy samples from the clean set and mitigate confirmation biasduring training. Experimental results on multiple benchmark datasetsdemonstrate the effectiveness of our proposed method in comparison with thestate-of-the-art approaches.

Learning with Noisy Labels Using Collaborative Sample Selection and Contrastive Semi-Supervised Learning