A Performance Prediction-based DNN Partitioner for Edge TPU Pipelining

Keyphrases

• Prediction Method 100%
• Performance Prediction 100%
• Partitioner 100%
• Tensor Processing Unit 100%
• Latency 66%
• DNN Model 66%
• On-chip Memory 33%
• Prediction Model 33%
• Low Power Consumption 33%
• Constrained Edge 33%
• High Computational Efficiency 33%
• Robust Performance 33%
• Tool Use 33%
• Memory Access 33%
• Resource-constrained 33%
• Mission-critical Applications 33%
• Edge Devices 33%
• Adversarial Environment 33%
• Memory Contention 33%
• Local Execution 33%
• Cloud Connection 33%
• Multiple Edges 33%
• Inference Time 33%
• Transformer Model 33%
• Workload Distribution 33%
• IoT Applications 33%
• Combined Prediction 33%
• Intelligent IoT 33%
• Edge TPU 33%
• AI pipeline 33%
• AI Hardware Accelerator 33%

Computer Science

• Pipelining 100%
• Processing Unit 100%
• Deep Neural Network 100%
• Artificial Intelligence 40%
• Computational Efficiency 20%
• Hardware Accelerator 20%
• Memory Access 20%
• Low Power Consumption 20%
• Critical Application 20%
• Prediction Model 20%
• Memory Contention 20%
• Workload Distribution 20%
• IoT Application 20%