Reproduce “Fast Perfect $L_2$ Sampling” 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
David P. Woodruff, Samson Zhou
The algorithm samples coordinates from a dynamic turnstile stream by squared frequency, avoiding explicit updates to polynomially many conceptual copies. It tracks the largest random weights and simulates the remaining tail using Gaussian stability. The theorem gives expected polylogarithmic update time, O(log³ n log(1/δ)) bits of space, conditional error O(n⁻ᶜ), and failure probability at most δ; these are theoretical guarantees.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.