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

Deep metric learning (DML) enables learning with less supervision through itsemphasis on the similarity structure of representations. There has been muchwork on improving generalization of DML in settings like zero-shot retrieval,but little is known about its implications for fairness. In this paper, we arethe first to evaluate state-of-the-art DML methods trained on imbalanced data,and to show the negative impact these representations have on minority subgroupperformance when used for downstream tasks. In this work, we first definefairness in DML through an analysis of three properties of the representationspace -- inter-class alignment, intra-class alignment, and uniformity -- andpropose finDML, the fairness in non-balanced DML benchmark to characterizerepresentation fairness. Utilizing finDML, we find bias in DML representationsto propagate to common downstream classification tasks. Surprisingly, this biasis propagated even when training data in the downstream task is re-balanced. Toaddress this problem, we present Partial Attribute De-correlation (PARADE) tode-correlate feature representations from sensitive attributes and reduceperformance gaps between subgroups in both embedding space and downstreammetrics.

Is Fairness Only Metric Deep? Evaluating and Addressing Subgroup Gaps in Deep Metric Learning