Attacking a Bayesian posterior with score matching is equivalent to minimizing an MMD, and the attack recovers the sufficient statistics of the training data.
Differentially private simple linear regression
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
stat.ML 1years
2025 1verdicts
CONDITIONAL 1representative citing papers
citing papers explorer
-
On Reconstructing Training Data From Bayesian Posteriors and Trained Models
Attacking a Bayesian posterior with score matching is equivalent to minimizing an MMD, and the attack recovers the sufficient statistics of the training data.