Leveraging Public Data for Practical Private Query Release

Authors: Terrance Liu, Giuseppe Vietri, Thomas Steinke, Jonathan Ullman, Steven Wu

ICML 2021 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We provide a theoretical analysis and an empirical evaluation on the American Community Survey (ACS) and ADULT datasets, which shows that our method outperforms stateof-the-art methods.
Researcher Affiliation Collaboration 1Carnegie Mellon University, Pittsburgh, PA, USA 2University of Minnesota, Minneapolis, MN, USA 3Google, Mountain View, CA, USA 4Northeastern University, Boston, MA, USA.
Pseudocode Yes Algorithm 1 PMWPub
Open Source Code No The paper does not contain an explicit statement or link indicating that the source code for the described methodology is publicly available.
Open Datasets Yes American Community Survey (ACS). We evaluate all algorithms on the 2018 American Community Survey (ACS), obtained from the IPUMS USA database (Ruggles et al., 2020)... ADULT. We evaluate algorithms on the ADULT dataset from the UCI machine learning dataset repository (Dua & Graff, 2017).
Dataset Splits No On the ACS dataset, we select hyperparameters for PMWPub using 5-run averages on the corresponding validation sets (treated as private) derived from the 2014 ACS release.
Hardware Specification Yes Experiments are conducted using a single core on an i5-4690K CPU (3.50GHz) machine.
Software Dependencies No The paper does not specify any software dependencies with version numbers (e.g., Python, PyTorch, specific libraries).
Experiment Setup Yes A list of hyperparameters is listed in Table 2 in the appendix.