Practical applications of synthetic data generation / by Lisa Pilgram and Khaled El Emam. : CS11-522/2024-1-1E-PDF

"Synthetic data generation (SDG) is increasingly applied across sectors for privacy-preserving data sharing, de-biasing and augmentation. Each use case requires a distinct set of evaluation metrics that must account for the stochasticity of the SDG process: membership and attribute disclosure vulnerability are critical for privacy; fidelity and downstream task utility apply even more broadly; and fairness and diversity are relevant for de-biasing and augmentation, respectively. Presenting accumulated evidence and through exemplar case studies, it is shown that SDG can perform well across many of these use cases and key learnings from our experiences with synthetic health data are shared"--Abstract, page [2].

Lien permanent pour cette publication :
publications.gc.ca/pub?id=9.965832&sl=1

• Format MARC XML
• Format MARC HTML

Demander des formats alternatifs