Reproduce “Hermite Polynomial Features for Private Data Generation” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Margarita Vinaroz, Mohammad-Amin Charusaie, Frederik Harder, Kamil Adamczewski, Mijung Park
DP-HP generates synthetic data by training a generator against a privatized summary of sensitive records, using ordered Hermite polynomial features instead of DP-MERF’s random Fourier features. The method also combines marginal comparisons with sampled-dimension joint comparisons. On discretized Adult inputs, its 3-way marginal error is 0.377, versus 0.480 for DP-MERF at ε=0.1; this result concerns marginals, not downstream classification.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.