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

In Multi-Label Learning (MLL), it is extremely challenging to accuratelyannotate every appearing object due to expensive costs and limited knowledge.When facing such a challenge, a more practical and cheaper alternative shouldbe Single Positive Multi-Label Learning (SPMLL), where only one positive labelneeds to be provided per sample. Existing SPMLL methods usually assume unknownlabels as negatives, which inevitably introduces false negatives as noisylabels. More seriously, Binary Cross Entropy (BCE) loss is often used fortraining, which is notoriously not robust to noisy labels. To mitigate thisissue, we customize an objective function for SPMLL by pushing only one pair oflabels apart each time to prevent the domination of negative labels, which isthe main culprit of fitting noisy labels in SPMLL. To further combat such noisylabels, we explore the high-rankness of label matrix, which can also push apartdifferent labels. By directly extending from SPMLL to MLL with full labels, aunified loss applicable to both settings is derived. Experiments on realdatasets demonstrate that the proposed loss not only performs more robustly tonoisy labels for SPMLL but also works well for full labels. Besides, weempirically discover that high-rankness can mitigate the dramatic performancedrop in SPMLL. Most surprisingly, even without any regularization or fine-tunedlabel correction, only adopting our loss defeats state-of-the-art SPMLL methodson CUB, a dataset that severely lacks labels.

Pushing One Pair of Labels Apart Each Time in Multi-Label Learning: From Single Positive to Full Labels