AI Paper Atlas

Research map / Federated learning and privacy / Differential privacy

Spectral Gradient Orthogonalization Improves Differentially Private Training at Scale

Sabari Shanmugam, Nick Barnes, Kerry Taylor

ECCV 2026 · Federated learning and privacy › Differential privacy · approach: DP-SGD refinements · 1 citations (OpenAlex)

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 changedWhereMeasuredEffect
add spectral orthogonalization (DP-Muon) in place of plain DP-SGD update at large batchDP-SGD on WRN-16-4 · CIFAR-10 (eps=4, B=4096)test accuracy (%): 39.3 -> 48.7Improved
add spectral orthogonalization at small batch size (low SNR)DP-SGD on WRN-16-4 · CIFAR-10 (eps=4, B=256)test accuracy (%): 39.4 -> 39.0No clear effect
add spectral orthogonalization at small batch on CIFAR-100DP-SGD on WRN-16-4 · CIFAR-100 (eps=4, B<=1024)test accuracy (%): 26.6 -> 22.0Hurt
add spectral orthogonalization at very large batch on CIFAR-100DP-SGD on WRN-16-4 · CIFAR-100 (eps=4, B=8192)test accuracy (%): 21.6 -> 25.8Improved
add spectral orthogonalization on SVHN with large batchDP-SGD on WRN-16-4 · SVHN (eps=4, B=4096)test accuracy (%): 20.3 -> 35.2Improved
add orthogonalization with larger model (WRN-28-10)DP-SGD on WRN-28-10 · CIFAR-10 (eps=4, B=4096)test accuracy (%): 34.0 -> 54.9Improved
add orthogonalization on ResNet-18DP-SGD on ResNet-18 · CIFAR-10 (eps=4, B=4096)test accuracy (%): +14.9%Improved
add orthogonalization to temporal filtering (DiSK-Muon vs DiSK-SGD)DiSK-SGD · CIFAR-10 (eps=4, B=4096)test accuracy (%): 39.9 -> 45.8Improved
add orthogonalization to DOPPLER low-pass filteringDOPPLER-SGD · CIFAR-10 (eps=4, B=4096)test accuracy (%): 32.3 -> 50.3Improved
add temporal Kalman filtering at small batchDP-SGD · CIFAR-10 (eps=4, B=256)test accuracy (%): 39.4 -> 49.0Improved
use magnitude-preserving scaling (DP-Muon-S) instead of pure orthogonalization at high SNR (ablation)DP-Muon · CIFAR-10 (eps=4, B=4096)test accuracy (%): 48.7 -> 46.3Hurt
add orthogonalization in full fine-tuning at large batchDP-SGD on ViT-Small · CIFAR-100 full fine-tuning (eps=4, B=4096)test accuracy (%): 79.5 -> 89.3Improved

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy