Private frequency estimation via projective geometry

Authors: Vitaly Feldman, Jelani Nelson, Huy Nguyen, Kunal Talwar

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

Reproducibility Variable Result LLM Response
Research Type Experimental Our empirical evaluation shows a speedup of over 50x over PI-RAPPOR while using approximately 75x less memory for practically relevant parameter settings.
Researcher Affiliation Collaboration 1Apple, Cupertino, CA, USA 2UC Berkeley, CA, USA 3Northeastern University, MA, USA.
Pseudocode No The paper describes algorithms mathematically and textually, particularly in Section 3 and 4, and mentions dynamic programming, but it does not include a dedicated 'Pseudocode' or 'Algorithm' block.
Open Source Code Yes code and details on how to run the experiments used to generate data and plots are in our public repository at https://github.com/minilek/ private_frequency_oracles/.
Open Datasets No The paper mentions performing experiments on 'simple synthetic data' but does not provide any access information, generation method, or citation for a publicly available dataset.
Dataset Splits No The paper does not explicitly provide information about training, validation, or test dataset splits. It mentions using 'simple synthetic data' but no specific partitioning details.
Hardware Specification Yes All experiments were run on a Dell Precision T3600 with six Intel 3.2 GHz Xeon E5-1650 cores running Ubuntu 20.04 LTS
Software Dependencies Yes We implemented all algorithms and ran experiments in C++, using the GNU C++ compiler version 9.3.0
Experiment Setup Yes We took ϵ = 5, a practically relevant setting, and n = 10,000, k = 3,307,948; this setting of n is smaller than one would see in practice, but the runtimes of the algorithms considered are all linear in n plus additional terms that depend on k, ϵ