Differentially Private Database Release via Kernel Mean Embeddings

Authors: Matej Balog, Ilya Tolstikhin, Bernhard Schölkopf

ICML 2018 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental as well as basic empirical illustrations of their performance on synthetic datasets
Researcher Affiliation Academia 1MPI-IS, T ubingen, Germany 2University of Cambridge, UK.
Pseudocode Yes Algorithm 1 Differentially private database release via a synthetic data subspace of the RKHS
Open Source Code Yes Code: https://github.com/matejbalog/RKHS-private-database/
Open Datasets No The paper uses synthetic datasets generated from a mixture of two Gaussians for its empirical illustrations, but does not provide concrete access information (link, DOI, formal citation) for a publicly available or open dataset.
Dataset Splits No The paper uses synthetic data for its experiments but does not provide specific dataset split information (percentages, counts, or detailed methodology) for training, validation, or testing its own models.
Hardware Specification No The paper does not provide specific hardware details (e.g., GPU/CPU models, memory amounts) used for running its experiments.
Software Dependencies No The paper does not provide specific ancillary software details, such as library names with version numbers, needed to replicate the experiments.
Experiment Setup Yes Appendix B (Experimental Details) specifies the generation of synthetic datasets, the use of a Gaussian kernel with bandwidth σ = 1, the number of private data points N = 100,000, and for Algorithm 2, the number of random features J = 2000.