Renyi Differential Privacy Mechanisms for Posterior Sampling

Authors: Joseph Geumlek, Shuang Song, Kamalika Chaudhuri

NeurIPS 2017 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental In this section, we present the experimental results for our proposed algorithms for both exponential family and GLMs. Our experimental design focuses on two goals first, analyzing the relationship between λ and ϵ in our privacy guarantees and second, exploring the privacy-utility trade-off of our proposed methods in relation to existing methods.
Researcher Affiliation Academia Joseph Geumlek University of California, San Diego jgeumlek@cs.ucsd.edu Shuang Song University of California, San Diego shs037@eng.ucsd.edu Kamalika Chaudhuri University of California, San Diego kamalika@cs.ucsd.edu
Pseudocode Yes Algorithm 1 Direct Posterior... Algorithm 2 Diffused Posterior... Algorithm 3 Concentrated Posterior... Algorithm 4 Concentrated Posterior... Algorithm 5 Diffuse Posterior
Open Source Code No The paper does not provide any explicit statements about releasing source code for the methodology or links to a code repository.
Open Datasets Yes We conduct Bayesian logistic regression on three real datasets: Abalone, Adult and MNIST.
Dataset Splits No The paper states '1/3 of the each dataset is used for testing, and the rest for training.' and provides specific training and test sample counts, but does not explicitly mention a separate validation split or cross-validation strategy.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details (e.g., library or solver names with version numbers).
Experiment Setup Yes For all algorithms, we use an original Gaussian prior with β = 10-3. The posterior sampling is done using slice sampling with 1000 burn-in samples. ... 500 iterations of binary search were used to select r and m when needed. ... For privacy parameters, we set λ = 1, 10, 100 and ϵ {e-5, e-4, . . . , e3}.