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

Differentially private (DP) synthetic data sets are a solution for sharingdata while preserving the privacy of individual data providers. Understandingthe effects of utilizing DP synthetic data in end-to-end machine learningpipelines impacts areas such as health care and humanitarian action, where datais scarce and regulated by restrictive privacy laws. In this work, weinvestigate the extent to which synthetic data can replace real, tabular datain machine learning pipelines and identify the most effective synthetic datageneration techniques for training and evaluating machine learning models. Weinvestigate the impacts of differentially private synthetic data on downstreamclassification tasks from the point of view of utility as well as fairness. Ouranalysis is comprehensive and includes representatives of the two main types ofsynthetic data generation algorithms: marginal-based and GAN-based. To the bestof our knowledge, our work is the first that: (i) proposes a training andevaluation framework that does not assume that real data is available fortesting the utility and fairness of machine learning models trained onsynthetic data; (ii) presents the most extensive analysis of synthetic data setgeneration algorithms in terms of utility and fairness when used for trainingmachine learning models; and (iii) encompasses several different definitions offairness. Our findings demonstrate that marginal-based synthetic datagenerators surpass GAN-based ones regarding model training utility for tabulardata. Indeed, we show that models trained using data generated bymarginal-based algorithms can exhibit similar utility to models trained usingreal data. Our analysis also reveals that the marginal-based synthetic datagenerator MWEM PGM can train models that simultaneously achieve utility andfairness characteristics close to those obtained by models trained with realdata.

Assessment of Differentially Private Synthetic Data for Utility and Fairness in End-to-End Machine Learning Pipelines for Tabular Data