Betty: Enabling Large-Scale GNN Training with Batch-Level Graph Partitioning

Keyphrases

• Large Graphs 100%
• Graph Partitioning 100%
• Neural Network Training 100%
• Level Diagrams 100%
• Graph Neural Network 100%
• Sampling Rate 40%
• Redundancy 40%
• Aggregator 40%
• Deep Aggregation 40%
• Mini 20%
• Single GPU 20%
• Memory Capacity 20%
• Memory Bottleneck 20%
• Memory-based 20%
• Graph Embedding 20%
• Graph-based 20%
• Memory Footprint 20%
• GPU Memory 20%
• Batch Size 20%
• Load Imbalance 20%
• Neural Network Performance 20%
• Scalable Graph Neural Network 20%
• DNN Training 20%

Computer Science

• Graph Partition 100%
• Neural Network Training 100%
• Graph Neural Network 100%
• Sampling Rate 33%
• Graphics Processing Unit 33%
• Memory Footprint 16%
• Memory Capacity 16%
• Network Performance 16%
• Load Imbalance 16%
• Deep Neural Network 16%