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

The existence of noisy labels in real-world data negatively impacts theperformance of deep learning models. Although much research effort has beendevoted to improving robustness to noisy labels in classification tasks, theproblem of noisy labels in deep metric learning (DML) remains open. In thispaper, we propose a noise-resistant training technique for DML, which we nameProbabilistic Ranking-based Instance Selection with Memory (PRISM). PRISMidentifies noisy data in a minibatch using average similarity against imagefeatures extracted by several previous versions of the neural network. Thesefeatures are stored in and retrieved from a memory bank. To alleviate the highcomputational cost brought by the memory bank, we introduce an accelerationmethod that replaces individual data points with the class centers. Inextensive comparisons with 12 existing approaches under both synthetic andreal-world label noise, PRISM demonstrates superior performance of up to 6.06%in Precision@1.

Noise-resistant Deep Metric Learning with Ranking-based Instance Selection