Mosaic-CNN: A combined two-step zero prediction approach to trade off accuracy and computation energy in convolutional neural networks

Keyphrases

• Computation Energy 100%
• Convolutional Neural Network 100%
• Prediction Approach 100%
• Classification Accuracy 33%
• Most Significant Bit 33%
• Convolutional Layer 33%
• Multiply-accumulate Operation 33%
• Energy Saving 16%
• Numerical Results 16%
• CMOS Process 16%
• Feature Map 16%
• Computational Scheme 16%
• 65nm CMOS 16%
• Least Significant Bit 16%
• Output Features 16%
• Rectified Linear Unit (ReLU) 16%
• Convolutional Neural Network Accelerator 16%
• Reconfigurable Accelerator 16%
• VGG16 16%

Mathematics

• Tradeoff 100%
• Convolutional Neural Network 100%
• Convolution 28%
• Convolutional Layer 28%
• Constant Value 14%

Computer Science

• Convolutional Neural Network 100%
• Classification Accuracy 28%
• Most Significant Bit 28%
• Convolutional Layer 28%
• Constant Value 14%
• Output Feature Map 14%
• VGG-16 14%
• Convolution Result 14%