Reproduce “Differentially Private Preconditioned Power Iteration Method for Principal Component Analysis that Adapts to the Coherence” 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
Duy V. Le, Ta Duy Nguyen, Alina Ene, Huy L. Nguyen
The authors present a differentially private algorithm for estimating PCA’s top-k subspace when each row represents one person and has norm at most 1. It rescales block power iteration before adding privacy noise, extending coherence-adaptive guarantees from one direction to rank k. Under an eigengap condition, its error bound depends on coherence and improves on Analyze Gauss when the product of k, sigma-k squared, and coherence squared is small.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.