Privacy-Preserving Machine Learning

This thesis presents a novel approach to differentially private data synthesis using variational autoencoders (VAEs). We develop a …