Increasing PE Utilization with a SW/HW Co-Design Technique for Sparse Convolutional Neural Networks

Keyphrases

• Convolution Neural Network 100%
• Design Techniques 100%
• Processing Engine 100%
• Compressed Format 100%
• Sparse Convolutional Neural Network 100%
• Energy Consumption 50%
• Pruning 50%
• Effective Techniques 50%
• Computation Time 50%
• Network Size 50%
• Data Compression 50%
• Compression System 50%
• Irregular Distribution 50%
• AlexNet 50%
• Neural Network Accelerator 50%
• Sparse Convolution 50%

Computer Science

• Convolutional Neural Network 100%
• Design Technique 100%
• Participatory Design 100%
• SIMD 66%
• Energy Consumption 33%
• Computation Time 33%
• Data Compression 33%
• Compression Scheme 33%
• Computer Hardware 33%