A Multicore GNN Training Accelerator

Keyphrases

• Multicore 100%
• Neural Network Training 100%
• Graph Neural Network 100%
• Training Accelerator 100%
• On chip 40%
• Low Energy 20%
• Real-world Problems 20%
• Optimal Performance 20%
• Superior Performance 20%
• Neural Network Model 20%
• Memory Access 20%
• Machine Learning Models 20%
• Feature Vector 20%
• Energy Efficiency Improvement 20%
• GPU-based 20%
• FPGA-based 20%
• Graph Structure 20%
• Access Pattern 20%
• Dynamic Caching 20%
• Graph Model 20%
• Segmentation Approach 20%
• Data Reuse 20%
• Neural Network Accelerator 20%
• DRAM Access 20%
• Data Access Latency 20%
• Vector Segmentation 20%

Computer Science

• Multicore 100%
• Graph Neural Network 100%
• Neural Network Training 100%
• Application Specific Integrated Circuit 40%
• Field Programmable Gate Arrays 40%
• Graphics Processing Unit 40%
• Data Access 20%
• Large Data Set 20%
• Neural Network Model 20%
• Superior Performance 20%
• Memory Access 20%
• Access Pattern 20%
• Real-World Problem 20%
• Data Reuse 20%
• Segmentation Approach 20%
• Energy Efficiency 20%
• Optimal Performance 20%
• Feature Vector 20%
• Structured Data 20%
• Machine Learning 20%
• Learning System 20%

Engineering

• Graphics Processing Unit 100%
• Field Programmable Gate Arrays 100%
• Application Specific Integrated Circuit 100%
• Neural Network Training 100%
• Energy Efficiency 50%
• Network Model 50%
• Optimal Performance 50%
• Memory Access 50%
• Feature Vector 50%
• Learning System 50%