Complement Sparsification: Low-Overhead Model Pruning for Federated Learning

Keyphrases

• Low Overhead 100%
• Sparsification 100%
• Federated Learning 100%
• Model Pruning 100%
• Sparse Model 40%
• Vanilla 30%
• Computation Overhead 20%
• Pruning Strategy 20%
• Bidirectional Communication 20%
• Privacy-preserving 10%
• Mobile Devices 10%
• Popular 10%
• Computational Effort 10%
• Data Distribution 10%
• Iterative Process 10%
• Resource-constrained 10%
• Learning Paradigm 10%
• Global Model 10%
• Model Accuracy 10%
• Model Performance 10%
• Best Model 10%
• Complementary and 10%
• Communication Overhead 10%
• Benchmark Dataset 10%
• Pruning 10%
• Substantial Reduction 10%
• Low Computation 10%
• Communication Activities 10%
• IoT Devices 10%
• Local Trends 10%
• Distributed Deep Learning 10%
• Dense Model 10%
• Pruned Models 10%

Engineering

• Deep Learning 100%
• Internet of Things 100%
• Iterative Process 100%
• Raw Data 100%

Computer Science

• Federated Learning 100%
• Pruning Mechanism 25%
• Privacy Preserving 12%
• Approximation (Algorithm) 12%
• Internet of Things 12%
• Deep Learning 12%
• Performance Model 12%
• Iterative Process 12%
• Data Distribution 12%
• Communication Overhead 12%
• Model Accuracy 12%
• Complement Addition 12%