Online Management for Edge-Cloud Collaborative Continuous Learning: A Two-timescale Approach

Keyphrases

• Two-time Scale 100%
• Continual Learning 100%
• Edge-cloud Collaborative 100%
• Online Management 100%
• Two-time-scale Approach 100%
• Training Model 50%
• Time Application 50%
• Deep Learning 50%
• Low Computation 50%
• Communication Cost 50%
• Error Bound 50%
• Geographical Location 50%
• Optimization Framework 50%
• Global Constraints 50%
• Distributed Deep Learning 50%
• Model Dynamics 50%
• Generated Data 50%
• Uncertain Future 50%
• Network Change 50%
• Deep Learning Training 50%
• Coupling Effect 50%
• Local Model 50%
• Data Heterogeneity 50%
• Model Convergence 50%
• Model Aggregation 50%
• Online Optimization 50%
• Lyapunov Optimization 50%
• Data Offloading 50%
• Online Offloading 50%
• Continuous Training 50%
• Computation Capability 50%
• Synchronization Frequency 50%
• Model Synchronization 50%
• Fine Time-scales 50%
• Dynamic Offloading 50%
• Compromise Model 50%
• Data Dynamics 50%
• Computational Node 50%
• Long-term Costs 50%

Computer Science

• Deep Learning Method 100%
• Data Stream 100%
• Synchronism 100%
• Large Data Set 50%
• Optimization Technique 50%
• Communication Cost 50%
• Real-Time Application 50%
• Optimization Framework 50%
• Global Constraint 50%
• Data Heterogeneity 50%
• Computation Node 50%
• Lyapunov Optimization 50%
• Computation Capability 50%