Reproduce “ABCDP: Approximate Bayesian Computation with Differential Privacy” 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
Mijung Park, Margarita Vinaroz, Wittawat Jitkrittum
ABCDP privatizes rejection approximate Bayesian computation for sensitive observations. It adds noise to distance and threshold comparisons and uses the Sparse Vector Technique, tying privacy cost to accepted samples rather than rejected candidates, assuming bounded sensitivity. A tuberculosis simulator experiment found mean absolute error fell from 4.51 to 2.10 as dataset size rose from 100 to 1000; these were not real patient data.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.