Modular Meta-Learning for Power Control via Random Edge Graph Neural Networks

Keyphrases

• Graph Neural Network 100%
• Power Control 100%
• Random Edge 100%
• Meta-learning 100%
• Edge Graph 100%
• Channel State Information 33%
• Joint Training 33%
• Power Control Policy 33%
• Wireless Networks 16%
• Control Problem 16%
• Transmit Power 16%
• Design Methodology 16%
• Numerical Results 16%
• Adaptation 16%
• Training Approaches 16%
• Policy-based 16%
• Neural Network Architecture 16%
• Network Topology 16%
• Learning Techniques 16%
• Data Availability 16%
• Channel Coefficients 16%
• Efficient Optimization 16%
• Limited Channel 16%
• Information Data 16%
• Black Box 16%
• Spatial Weights 16%
• Stochastic Gradient Descent 16%
• Training Scheme 16%
• Graph Convolutional Network 16%
• Time-varying Topology 16%
• Linear Graphs 16%
• Policy Mapping 16%
• Data-driven Design 16%
• Adaptation Procedure 16%
• Convolutional Filter 16%
• Reusable Modules 16%

Engineering

• Power Control 100%
• Joints (Structural Components) 66%
• Channel State Information 66%
• Power Control Policy 66%
• Nodes 33%
• Learning Technique 33%
• Network Topology 33%
• Neural Network Architecture 33%
• Black Box 33%
• Gradient Descent 33%
• Component Form 33%

Computer Science

• Graph Neural Network 100%
• Channel State Information 50%
• Wireless Networks 25%
• Network Topology 25%
• Neural Network Architecture 25%
• Learning Technique 25%
• Data Availability 25%
• Gradient Descent 25%
• Convolutional Filter 25%