A Joint Exponential Mechanism For Differentially Private Top-$k$

Authors: Jennifer Gillenwater, Matthew Joseph, Andres Munoz, Monica Ribero Diaz

ICML 2022 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments compare the peeling and joint mechanisms across several real-world datasets using the error metrics from Definition 2.2.
Researcher Affiliation Collaboration 1Google Research NYC 2UT Austin.
Pseudocode Yes Algorithm 1 Efficiently sampling JOINT
Open Source Code Yes All datasets and experiment code are public (Google, 2022).
Open Datasets Yes We use six datasets: Books (Soumik, 2019) (11,000+ Goodreads books with review counts), Foods (Mc Auley, 2014) (166,000+ Amazon foods with review counts), Games (Tamber, 2016) (5,000+ Steam games with purchase counts), Movies (Harper & Konstan, 2015) (62,000+ Movies with rating counts), News (Fernandes et al., 2015) (40,000+ Mashable articles with share counts), and Tweets (Bin Tareaf, 2017) (52,000+ Tweets with like counts).
Dataset Splits No The paper does not provide specific dataset split information (exact percentages, sample counts, citations to predefined splits, or detailed splitting methodology) for training, validation, or testing.
Hardware Specification No The paper does not provide specific hardware details (exact GPU/CPU models, processor types with speeds, memory amounts, or detailed computer specifications) 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) needed to replicate the experiment.
Experiment Setup Yes We use k = 5, 15, . . . , 195 with 1-DP instances of JOINT and PNF-PEEL and (1, 10 6)-DP instances of CDP-PEEL.