Optimally removing synchronization overhead for CNNs in three-dimensional neuromorphic architecture

Keyphrases

• Three-dimensional (3D) 100%
• Synchronization Overheads 100%
• Neuromorphic Architecture 100%
• Processing-In-Memory Architectures 66%
• Processing Time 33%
• Limited Capacity 33%
• Computational Resources 33%
• Target Problem 33%
• Schedule Length 33%
• Neuromorphic Computing 33%
• Convolutional Neural Network 33%
• Memory Resource 33%
• Deep Learning Framework 33%
• Neural Network Applications 33%
• Task Schedule 33%
• Dynamic Programming Model 33%
• Caffe 33%
• TensorFlow 33%
• Computational Tasks 33%
• Hardware Efficiency 33%

Computer Science

• Convolutional Neural Network 100%
• Memory Architecture 100%
• Programming Model 50%
• Deep Learning Method 50%
• Learning Framework 50%
• Speed-up 50%
• Dynamic Programming 50%
• Computer Hardware 50%
• Memory Resource 50%
• Neural Network Application 50%
• Processing Core 50%
• Processing Time 50%
• TensorFlow 50%

Engineering

• Convolutional Neural Network 100%
• Network Application 50%
• Programming Model 50%
• Dynamic Programming 50%
• Processing Time 50%
• Deep Learning Method 50%