StarSPA: Stride-Aware Sparsity Compression for Efficient CNN Acceleration

Keyphrases

• Sparsity 100%
• Efficient CNN 100%
• CNN Acceleration 100%
• Compression Method 50%
• Sparse Input 50%
• Overall Power 25%
• Energy Efficiency 25%
• Power Efficiency 25%
• Power Consumption 25%
• Hardware Overhead 25%
• Comprehensive Simulation 25%
• Input Values 25%
• Low Quality Data 25%
• Convolutional 25%
• Matching Process 25%
• Storage Requirement 25%
• FPGA Implementation 25%
• Data Compression 25%
• Data Transfer 25%
• Input Matching 25%
• Convolutional Neural Network 25%
• GoogLeNet 25%
• AlexNet 25%
• Factorized Convolution 25%
• VGG16 25%
• Squeeze-and-excitation Network (SENet) 25%
• EfficientNet 25%
• Transfer Price 25%
• Innovative Data 25%
• Speedup Ratio 25%

Engineering

• Sparsity 100%
• Convolutional Neural Network 100%
• Compression Method 50%
• Energy Conservation 25%
• Input Feature 25%
• Electric Power Utilization 25%
• Energy Efficiency 25%
• Field Programmable Gate Arrays 25%
• Input Data 25%
• Hardware Overhead 25%
• Storage Requirement 25%
• Matching Process 25%
• Data Transfer 25%
• Input Value 25%

Computer Science

• Sparsity 100%
• Convolutional Neural Network 100%
• Power Consumption 25%
• Energy Efficiency 25%
• Power Efficiency 25%
• Hardware Overhead 25%
• Storage Requirement 25%
• Matching Process 25%
• Data Compression 25%
• Multiplication Process 25%
• Computer Hardware 25%
• Field Programmable Gate Arrays 25%