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/

Zendy Plus

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

Zendy Tools

Zendy Open

Long-tailed semi-supervised learning (LTSSL) represents a practical scenariofor semi-supervised applications, challenged by skewed labeled distributionsthat bias classifiers. This problem is often aggravated by discrepanciesbetween labeled and unlabeled class distributions, leading to biasedpseudo-labels, neglect of rare classes, and poorly calibrated probabilities. Toaddress these issues, we introduce Flexible Distribution Alignment (FlexDA), anovel adaptive logit-adjusted loss framework designed to dynamically estimateand align predictions with the actual distribution of unlabeled data andachieve a balanced classifier by the end of training. FlexDA is furtherenhanced by a distillation-based consistency loss, promoting fair data usageacross classes and effectively leveraging underconfident samples. This method,encapsulated in ADELLO (Align and Distill Everything All at Once), provesrobust against label shift, significantly improves model calibration in LTSSLcontexts, and surpasses previous state-of-of-art approaches across multiplebenchmarks, including CIFAR100-LT, STL10-LT, and ImageNet127, addressing classimbalance challenges in semi-supervised learning. Our code is available athttps://github.com/emasa/ADELLO-LTSSL.

Flexible Distribution Alignment: Towards Long-tailed Semi-supervised Learning with Proper Calibration