TAIT: One-Shot Full-Integer Lightweight DNN Quantization via Tunable Activation Imbalance Transfer

Mathematics

• Activation 100%
• Quantization 82%
• Neural Networks 79%
• Field Programmable Gate Array 44%
• Integer 44%
• Uniformity 29%
• Predictors 26%
• Energy Efficiency 23%
• Convolution 23%
• Tensor 22%
• Accelerator 19%
• Scaling Factor 18%
• System Design 18%
• Range of data 17%
• Guidance 16%
• Framework 16%
• High Accuracy 14%
• Resources 13%
• Requirements 11%
• Methodology 10%
• Demonstrate 8%

Engineering & Materials Science

• Deep neural networks 82%
• Chemical activation 75%
• Tensors 42%
• Convolution 41%
• Field programmable gate arrays (FPGA) 34%
• Application specific integrated circuits 22%
• Particle accelerators 21%
• Energy efficiency 16%
• Systems analysis 15%