/lsq

Differentially private kernel density estimation (DP-KDE) via Locality Sensitive Quantization (LSQ)

Primary LanguagePythonMIT LicenseMIT

Differentially private kernel density estimation (DP-KDE) via Locality Sensitive Quantization (LSQ)

This is an accompanying implementation for the paper: Fast Private Kernel Density Estimation via Locality Sensitive Quantization, by Tal Wagner, Yonatan Naamad and Nina Mishra, published in ICML 2023.

The code implements the LSQ-RFF and LSQ-FGT mechanisms for the Gaussian kernel.