BD-NET: A multiplication-less DNN with binarized depthwise separable convolution

Keyphrases

• Binarization 100%
• Depthwise Separable Convolution 100%
• Convolutional Layer 66%
• Deep Convolutional Neural Network (deep CNN) 66%
• Hardware Efficiency 33%
• Hardware Implementation 33%
• Computation Cost 33%
• Subtraction by Addition 33%
• Convolution Neural Network 33%
• Memory Consumption 33%
• Memory chip 33%
• Experiment Results 33%
• Subtraction Technique 33%
• Computation Energy 33%
• Inference Accuracy 33%
• Network Configuration 33%
• Drop-in Replacement 33%
• Chip Area 33%
• CIFAR10 Dataset 33%
• Convolution Operation 33%
• Accuracy Parameters 33%
• Spatial Convolution 33%
• Convolution Block 33%
• Multiplication-and-accumulation 33%
• Usage Areas 33%

Computer Science

• Deep Neural Network 100%
• Convolutional Neural Network 100%
• Depthwise Separable Convolution 100%
• Convolution Layer 66%
• Hardware Implementation 33%
• Computation Cost 33%
• Network Configuration 33%
• Parameter Size 33%
• Convolution Operation 33%
• Computer Hardware 33%