AI Paper Atlas

Research map / Federated learning and privacy / Differential privacy

Initialization Matters: Privacy-Utility Analysis of Overparameterized Neural Networks

Jiayuan Ye, Zhenyu Zhu, Fanghui Liu, Reza Shokri, Volkan Cevher

NeurIPS 2023 · Federated learning and privacy › Differential privacy · approach: Private online and user-level learning

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
Increase widthFully connected ReLU networks trained with noisy gradient descent · Privacy vs width, depth 10, 50 epochsKL privacy lossHurt
Increase depth with LeCun initializationFully connected ReLU networks trained with noisy gradient descent · Privacy vs depth, width 1024, 20 epochsKL privacy lossMixed
Increase depth with Xavier initializationFully connected ReLU networks trained with noisy gradient descent · Privacy vs depth, width 1024, 20 epochsKL privacy lossMixed
Increase depth with He initializationFully connected ReLU networks trained with noisy gradient descent · Privacy vs depth, width 1024, 20 epochsKL privacy lossHurt
Increase depthFully connected ReLU networks trained with noisy gradient descent · Privacy vs depth, width 1024, 50 epochsKL privacy lossHurt
Use LeCun initialization rather than an initialization with larger per-layer varianceLinearized fully connected ReLU network · KL privacy boundgradient norm constant BImproved
Increase depth under LeCun initializationLinearized fully connected ReLU network · KL privacy boundKL privacyImproved
Increase depth under Xavier initializationLinearized fully connected ReLU network · KL privacy boundKL privacyImproved
Increase depth under He initializationLinearized fully connected ReLU network · KL privacy boundKL privacyHurt
Increase depth under NTK initializationLinearized fully connected ReLU network · KL privacy boundKL privacyHurt
Increase depth under LeCun initializationLinearized fully connected ReLU network · Excess empirical risk under fixed KL privacyexcess empirical riskImproved
Increase depth under Xavier initializationLinearized fully connected ReLU network · Excess empirical risk under fixed KL privacyexcess empirical riskImproved

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