Research map / Federated learning and privacy / Differential privacy
Shape And Structure Preserving Differential Privacy
Carlos J Soto, Karthik Bharath, Matthew Reimherr, Aleksandra Slavkovic
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| Use manifold KNG instead of manifold Laplace | Fréchet mean sanitization on the unit sphere · Unit sphere S² | average Euclidean distance | Improved |
| Use manifold KNG instead of Euclidean Laplace followed by projection | Fréchet mean sanitization on the unit sphere · Unit sphere S² | average Euclidean distance | Improved |
| Project Euclidean Laplace output onto the sphere (ablation) | Euclidean Laplace for Fréchet mean sanitization · Unit sphere S² | average Euclidean distance | Improved |
| Use manifold KNG instead of Euclidean Laplace | Fréchet mean sanitization on SPD matrices · SPD matrices P(2) | average Euclidean distance | Improved |
| Use manifold KNG instead of manifold Laplace | Fréchet mean sanitization on SPD matrices · SPD matrices P(2) | average Euclidean distance | No clear effect |
| Sanitize the mean on Kendall shape space rather than point-wise with Laplace after rotational alignment | Corpus callosum mean shape · ADNI corpus callosa | shape structure preservation | Improved |
| Account for rotational alignment before point-wise Laplace sanitization (ablation) | Point-wise Laplace sanitization of corpus callosum mean · ADNI corpus callosa | shape structure preservation | Improved |
| Use manifold KNG rather than rotationally aligned point-wise Laplace | Smoothed private corpus callosum estimates · ADNI corpus callosa | shape smoothness and structural distortion | Improved |
Closest papers by meaning
- Differential Privacy Over Riemannian Manifolds (NeurIPS 2021 · similarity 0.93)
- Exponential-Wrapped Mechanisms: Differential Privacy on Hadamard Manifolds Made Practical (ICLR 2026 · similarity 0.89)
- Differentially Private Geodesic Regression (ICML 2026 · similarity 0.88)
- Wrapped Gaussian on the manifold of Symmetric Positive Definite Matrices (ICML 2025 · similarity 0.85)
- Intrinsic Riemannian Cross-covariance for Manifold-valued Random Objects (NeurIPS 2026 · similarity 0.85)
- Gaussian Differentially Private Human Faces Under a Face Radial Curve Representation (ICLR 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
