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

Electronic Health Records (EHRs) are commonly used by the machine learningcommunity for research on problems specifically related to health care andmedicine. EHRs have the advantages that they can be easily distributed andcontain many features useful for e.g. classification problems. What makes EHRdata sets different from typical machine learning data sets is that they areoften very sparse, due to their high dimensionality, and often containheterogeneous (mixed) data types. Furthermore, the data sets deal withsensitive information, which limits the distribution of any models learnedusing them, due to privacy concerns. For these reasons, using EHR data inpractice presents a real challenge. In this work, we explore using GenerativeAdversarial Networks to generate synthetic, heterogeneous EHRs with the goal ofusing these synthetic records in place of existing data sets for downstreamclassification tasks. We will further explore applying differential privacy(DP) preserving optimization in order to produce DP synthetic EHR data sets,which provide rigorous privacy guarantees, and are therefore shareable andusable in the real world. The performance (measured by AUROC, AUPRC andaccuracy) of our model's synthetic, heterogeneous data is very close to theoriginal data set (within 3 - 5% of the baseline) for the non-DP model whentested in a binary classification task. Using strong $(1, 10^{-5})$ DP, ourmodel still produces data useful for machine learning tasks, albeit incurring aroughly 17% performance penalty in our tested classification task. Weadditionally perform a sub-population analysis and find that our model does notintroduce any bias into the synthetic EHR data compared to the baseline ineither male/female populations, or the 0-18, 19-50 and 51+ age groups in termsof classification performance for either the non-DP or DP variant.

Generation of Differentially Private Heterogeneous Electronic Health Records