Understanding Clipping for Federated Learning: Convergence and Client-Level Differential Privacy

Keyphrases

• Federated Learning 100%
• Learning Convergence 100%
• Client-level Differential Privacy 100%
• Learning Algorithm 40%
• Differentially Private 40%
• FedAvg 40%
• Motivation 20%
• Differential Privacy 20%
• Neural Network Training 20%
• Convergence Analysis 20%
• Model Update 20%
• Privacy Protection 20%
• Data Heterogeneity 20%
• Deep Learning Architectures 20%
• Privacy Notions 20%
• Popular 20%
• Gradient Clipping 20%
• Formal Privacy 20%

Computer Science

• Differential Privacy 100%
• Federated machine learning 100%
• Learning Algorithm 40%
• Mathematical Convergence 20%
• Privacy Protection 20%
• Data Heterogeneity 20%
• Deep Architecture 20%
• Neural Network Training 20%

Mathematics

• Federated machine learning 100%
• Convergence Analysis 20%
• Neural Network 20%