Convolution Filter Compression via Sparse Linear Combinations of Quantized Basis

Keyphrases

• Linear Combination 100%
• Convolution Filters 100%
• Filter Method 66%
• Convolutional Neural Network 66%
• Sparsity 33%
• Image Classification 33%
• Image Object 33%
• Network Structure 33%
• Existing State 33%
• Object Detection 33%
• Computational Resources 33%
• Comparable Accuracy 33%
• Number of Parameters 33%
• Classification Detection 33%
• Convolutional Layer 33%
• High Compression Ratio 33%
• Compression Method 33%
• Overfitting 33%
• Real-time Task 33%
• Storage Resources 33%
• Detection Task 33%
• Quantization Method 33%
• Embedded Devices 33%
• Compression Performance 33%
• Storage Utilization 33%
• Memory-constrained 33%
• Filter Network 33%
• Resource Making 33%
• Filter Decomposition 33%
• Compact Filter 33%
• Neural Network Quantization 33%

Computer Science

• Convolution Filter 100%
• Sparse Linear Combination 100%
• Convolutional Neural Network 66%
• Linear Combination 66%
• Image Classification 33%
• Sparsity 33%
• Compression Ratio 33%
• Network Structure 33%
• Object Detection 33%
• Convolutional Layer 33%
• Embedded Device 33%
• Compression Performance 33%