Reproduce “BRACE: Differential Privacy for Dense Associative Memory with LSR Energy” 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
Chang Qu, Zhaoyang Shi
BRACE provides differential privacy for retrieval in log-sum-ReLU dense associative memory, where small noise can change which records fall within a fixed radius. It detects and corrects discrepancies between clean and noisy retrievals before adding calibrated privacy noise. On MNIST, its best prediction MSE is 1.582963, tying Naive DAM in a two-coordinate feature space.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.