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 Tools annual

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 Tools monthly

Global: Zendy Plus monthly

Presents the access of premium content as premium feature

Premium Content

Global: Zendy Plus annual

Global: Zendy Free Plan

Deep learning is becoming increasingly ubiquitous in medical research andapplications while involving sensitive information and even critical diagnosisdecisions. Researchers observe a significant performance disparity amongsubgroups with different demographic attributes, which is called modelunfairness, and put lots of effort into carefully designing elegantarchitectures to address unfairness, which poses heavy training burden, bringspoor generalization, and reveals the trade-off between model performance andfairness. To tackle these issues, we propose FairAdaBN by making batchnormalization adaptive to sensitive attribute. This simple but effective designcan be adopted to several classification backbones that are originally unawareof fairness. Additionally, we derive a novel loss function that restrainsstatistical parity between subgroups on mini-batches, encouraging the model toconverge with considerable fairness. In order to evaluate the trade-off betweenmodel performance and fairness, we propose a new metric, namedFairness-Accuracy Trade-off Efficiency (FATE), to compute normalized fairnessimprovement over accuracy drop. Experiments on two dermatological datasets showthat our proposed method outperforms other methods on fairness criteria andFATE.

FairAdaBN: Mitigating unfairness with adaptive batch normalization and its application to dermatological disease classification