Reproduce “DP-LDMs: Differentially Private Latent Diffusion Models” 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
Michael F. Liu, Saiyue Lyu, Margarita Vinaroz, Mijung Park
DP-LDM adapts pretrained diffusion models to private image data by encoding images in a lower-dimensional latent space and using DP-SGD to update selected attention modules. On CIFAR-10 at ε=10 and δ=10⁻⁵, it reports FID 8.4 ± 0.2, compared with 9.8 for DP-Diffusion. It also generates text-conditioned 256×256 faces, but privacy attacks were not evaluated.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.