rization for scalable and privacy-preserving Bayesian inference in high-dimensions is recognized as important for the