Private Counting from Anonymous Messages: Near-Optimal Accuracy with Vanishing Communication Overhead

Authors: Badih Ghazi, Ravi Kumar, Pasin Manurangsi, Rasmus Pagh

ICML 2020 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental We implement our algorithms and compare their performance to several alternatives proposed in the literature, both for binary summation and histogram.
Researcher Affiliation Collaboration 1Google Research, Mountain View 2IT University of Copenhagen.
Pseudocode Yes Algorithm 1 D-Distributed Randomizer. Algorithm 2 D-Distributed Analyzer. Algorithm 3 (D1, D2, D3)-Correlated Distributed Randomizer. Algorithm 4 (D1, D2, D3)-Correlated Distributed Analyzer.
Open Source Code No The paper mentions 'The experimental results are presented in the supplementary material' but does not explicitly state that the source code for the methodology is released or provide a link.
Open Datasets Yes For the latter, we evaluate the algorithms on public 1940 US Census IPUMS dataset, considering both categorical and numerical features. We ran these algorithms on two IPUMS datasets (Ruggles et al., 2019).
Dataset Splits No The paper mentions evaluating algorithms on datasets but does not explicitly provide specific details about training, validation, or test dataset splits in the main text.
Hardware Specification No The paper does not provide specific hardware details (e.g., CPU, GPU models, memory, or cloud instance types) used for running its experiments.
Software Dependencies No The paper discusses implementing algorithms and comparing them to other methods, but it does not specify any software dependencies with version numbers.
Experiment Setup No The paper states, 'For each setting of ε, δ, our parameters are selected in an accurate manner; this is explained in detail in the supplementary material,' indicating that specific experimental setup details are not in the main text.