Minimizing area and energy of deep learning hardware design using collective low precision and structured compression

Mathematics

• Hardware Design 100%
• Compression 59%
• Neural Networks 54%
• Learning 52%
• Energy 38%
• Sparsity 36%
• Hardware 28%
• Learning Algorithm 27%
• Binarization 19%
• Pruning 14%
• Network Structure 14%
• Floating point 14%
• Neural Network Model 14%
• Degradation 14%
• Activation 13%
• Training 12%
• Regularization 10%
• Design 7%
• Performance 6%

Engineering & Materials Science

• Deep neural networks 56%
• Deep learning 51%
• Computer hardware 34%
• Learning algorithms 19%
• Data storage equipment 14%
• Chemical activation 10%
• Degradation 8%
• Neural networks 7%

Physics & Astronomy

• learning 63%
• hardware 55%
• energy 17%
• footprints 9%
• floating 8%
• education 8%
• CMOS 7%
• platforms 6%
• activation 6%
• performance 3%