FedAT: A High-Performance and Communication-Efficient Federated Learning System with Asynchronous Tiers

Keyphrases

• High Performance 100%
• Federated Learning 100%
• Communication Cost 100%
• Efficient Federated Learning Systems 100%
• Communication-efficient Federated Learning 100%
• Communication Bottleneck 66%
• Convergence Rate 66%
• Training Data 66%
• Learning Methods 66%
• Straggler Problem 66%
• Model Accuracy 33%
• Collaborative Learning 33%
• Accuracy Improvement 33%
• Performance Prediction 33%
• Central Server 33%
• Response Latency 33%
• Test Accuracy 33%
• Single-dimension 33%
• Synchronous and Asynchronous 33%
• Central Bottleneck 33%
• Convergence Test 33%
• Network Resources 33%
• Improved Convergence 33%
• Compression Algorithms 33%
• Asynchronous Methods 33%
• Stragglers 33%
• Speed-accuracy 33%
• Model Updating 33%
• Polyline 33%
• Resource Heterogeneity 33%
• Asynchronous Training 33%
• Local Update 33%
• Asynchronous Model 33%
• Weighted Aggregation 33%
• Speed Model 33%
• Distributed Devices 33%
• Synchronous Training 33%
• Model Convergence 33%
• Non-IID 33%
• Computing Resources 33%
• Federated Learning Framework 33%
• Uplink Communication 33%
• Straggler Effect 33%
• Synchronous Mechanism 33%
• Downlink Communication 33%

Computer Science

• Learning System 100%
• Federated machine learning 100%
• Communication Cost 60%
• Training Data 40%
• Convergence Speed 40%
• Computing Resource 20%
• Model Accuracy 20%
• Introduce Bias 20%
• Network Resource 20%
• Prediction Performance 20%
• Compression Algorithm 20%
• Asynchronous Method 20%
• Collaborative Learning 20%