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

High model performance, on average, can hide that models may systematicallyunderperform on subgroups of the data. We consider the tabular setting, whichsurfaces the unique issue of outcome heterogeneity - this is prevalent in areassuch as healthcare, where patients with similar features can have differentoutcomes, thus making reliable predictions challenging. To tackle this, wepropose Data-IQ, a framework to systematically stratify examples into subgroupswith respect to their outcomes. We do this by analyzing the behavior ofindividual examples during training, based on their predictive confidence and,importantly, the aleatoric (data) uncertainty. Capturing the aleatoricuncertainty permits a principled characterization and then subsequentstratification of data examples into three distinct subgroups (Easy, Ambiguous,Hard). We experimentally demonstrate the benefits of Data-IQ on four real-worldmedical datasets. We show that Data-IQ's characterization of examples is mostrobust to variation across similarly performant (yet different) models,compared to baselines. Since Data-IQ can be used with any ML model (includingneural networks, gradient boosting etc.), this property ensures consistency ofdata characterization, while allowing flexible model selection. Taking this astep further, we demonstrate that the subgroups enable us to construct newapproaches to both feature acquisition and dataset selection. Furthermore, wehighlight how the subgroups can inform reliable model usage, noting thesignificant impact of the Ambiguous subgroup on model generalization.

Data-IQ: Characterizing subgroups with heterogeneous outcomes in tabular data