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

Removing bias while keeping all task-relevant information is challenging forfair representation learning methods since they would yield random ordegenerate representations w.r.t. labels when the sensitive attributescorrelate with labels. Existing works proposed to inject the label informationinto the learning procedure to overcome such issues. However, the assumptionthat the observed labels are clean is not always met. In fact, label bias isacknowledged as the primary source inducing discrimination. In other words, thefair pre-processing methods ignore the discrimination encoded in the labelseither during the learning procedure or the evaluation stage. Thiscontradiction puts a question mark on the fairness of the learnedrepresentations. To circumvent this issue, we explore the following question:\emph{Can we learn fair representations predictable to latent ideal fair labelsgiven only access to unreliable labels?} In this work, we propose a\textbf{D}e-\textbf{B}iased \textbf{R}epresentation Learning for\textbf{F}airness (DBRF) framework which disentangles the sensitive informationfrom non-sensitive attributes whilst keeping the learned representationspredictable to ideal fair labels rather than observed biased ones. We formulatethe de-biased learning framework through information-theoretic concepts such asmutual information and information bottleneck. The core concept is that DBRFadvocates not to use unreliable labels for supervision when sensitiveinformation benefits the prediction of unreliable labels. Experiment resultsover both synthetic and real-world data demonstrate that DBRF effectivelylearns de-biased representations towards ideal labels.

De-biased Representation Learning for Fairness with Unreliable Labels