Fast Private Kernel Density Estimation via Locality Sensitive Quantization

Authors: Tal Wagner, Yonatan Naamad, Nina Mishra

ICML 2023 | Conference PDF | Archive PDF | Plain Text | LLM Run Details

Reproducibility Variable Result LLM Response
Research Type Experimental Our experiments demonstrate that our resulting DP-KDE mechanisms are fast and accurate on large datasets in both high and low dimensions.
Researcher Affiliation Industry 1Amazon. Correspondence to: Tal Wagner <tal.wagner@gmail.com>.
Pseudocode Yes Algorithm 1: LSQ Mechanism for DP-KDE Curator
Open Source Code Yes Our code is available online.2 (2https://github.com/talwagner/lsq)
Open Datasets Yes Covertype: forest cover types (n = 581,012, d = 55) (Blackard & Dean, 1999) Glo Ve: word embeddings (n = 1,000,000, d = 100) (Pennington et al., 2014) Diabetes: age and days in hospital (n = 101,766, d = 2) (Strack et al., 2014) NYC Taxi: longitude and latitude (n = 100,000, d = 2) (Chavez et al., 2018)
Dataset Splits No The paper mentions holding out query points but does not specify explicit training/validation/test splits (e.g., percentages, counts, or cross-validation).
Hardware Specification No The paper does not explicitly describe the hardware (e.g., specific GPU/CPU models, memory, or cloud instances) used for running experiments.
Software Dependencies No The paper does not specify any software dependencies with version numbers (e.g., Python 3.8, PyTorch 1.9).
Experiment Setup Yes In LSQ-RFF, we parameterize the mechanism by the number of random Fourier features... In LSQ-FGT, the user selects an integer parameter ρ ≥ 1... For each dataset we tune the bandwidth according to the guidelines in prior work... Bandwidth values are tuned are such that mean KDE values are on the order of 10^-2 and their standard deviation is also on the order of 10^-2