VecQ: Minimal Loss DNN Model Compression with Vectorized Weight Quantization

Keyphrases

• Model Compression 100%
• Vectorized 100%
• Quantization Loss 100%
• DNN Model 100%
• Weight Quantization 100%
• Model Accuracy 40%
• Size Reduction 20%
• Process Optimization 20%
• Independent Optimization 20%
• Effective Method 20%
• Estimation Method 20%
• New Metric 20%
• Nonconvex 20%
• Feature Extraction 20%
• Loss Minimization 20%
• Detection Task 20%
• Control Accuracy 20%
• Minimization Problem 20%
• Template-based 20%
• Training Model 20%
• Text Data 20%
• Novel Metrics 20%
• High Loss 20%
• ImageNet 20%
• Storage Cost 20%
• Probability Estimation 20%
• Movie Reviews 20%
• MNIST 20%
• Local Optimization Method 20%
• Saliency Detection 20%
• L2 Loss 20%
• DNN Quantization 20%
• CIFAR 20%
• Vector Loss 20%
• Quantized Model 20%

Computer Science

• Deep Neural Network 100%
• Model Compression 100%
• Model Accuracy 33%
• Quantization Process 33%
• Effective Method 16%
• Estimation Method 16%
• Speed-up 16%
• Training Model 16%
• Loss Minimization 16%
• Process Optimization 16%
• Local Optimization 16%
• Independent Optimization 16%
• Probability Estimation 16%
• Salient Object Detection 16%
• Feature Extraction 16%