Open vollmersj opened 3 years ago
@vollmersj Please have a look at this (and the developer's answer). I created a forked version that allows the user to define their own Bayesian Network structure and pass it to DataSynthesizer. This solves the problem I described but obviously that part of the synthesis process is no longer differentially private as the developer points out.
If you have one step which is non-DP and all the other steps adhering to DP, is this still a method that can be considered DP (but without any protection for the network structure)? Or is it not DP at all and thus not comparable with other DP methods?
it would be great if we can support https://github.com/DataResponsibly/DataSynthesizer