Compressing Deep Convolutional Neural Networks by Stacking Low-dimensional Binary Convolution Filters

Keyphrases

• Benchmark Dataset 16%
• Binary Convolution 100%
• Binary Convolutional Neural Network 50%
• Binary Filter 16%
• Comparable Accuracy 16%
• Compress 16%
• Compression Ratio 33%
• Compression Techniques 16%
• Computation Cost 16%
• Convolution Filters 100%
• Convolutional Layer 16%
• Convolutional Neural Network Model 33%
• Deep Convolutional Neural Network (deep CNN) 100%
• Feature Map 16%
• High Compression Ratio 16%
• Intermediate Features 16%
• Memory Cost 50%
• Memory-Constrained Devices 16%
• Mobile Phone 16%
• Model Compression 16%
• Model Inference 16%
• Popular 33%
• Real-world Problems 16%
• Single-bit 16%
• Split-transform-merge 16%
• VGG-16 Network 16%

Computer Science

• Compression Ratio 50%
• Compression Technique 16%
• Computation Cost 16%
• Constrained Device 16%
• Convolution Filter 100%
• Convolution Layer 16%
• Convolutional Neural Network 50%
• Deep Convolutional Neural Networks 100%
• Experimental Result 16%
• Feature Map 16%
• Inference Model 16%
• Model Compression 16%
• Neural Network Model 83%
• VGG-16 16%

Mathematics

• Approximates 12%
• Compression Ratio 37%
• Convolution 100%
• Convolutional Neural Network 100%
• Network Model 62%
• Real Life 12%
• Single Bit 12%