HW-FlowQ: A Multi-Abstraction Level HW-CNN Co-design Quantization Methodology

Keyphrases

• Convolutional Neural Network 100%
• Levels of Abstraction 100%
• Multi-Abstraction 100%
• Search Space 33%
• Design Space 33%
• Convolutional Neural Network Model 16%
• Latency 16%
• Model Compression 16%
• Iterative Approach 16%
• Solution Space 16%
• Prediction Accuracy 16%
• Hardware Design 16%
• Search Methods 16%
• Pareto Optimal Set 16%
• Bit-parallel 16%
• Bit-serial 16%
• Processing Unit 16%
• Embedded Platform 16%
• Hardware Architecture 16%
• Hardware Platform 16%
• Modeling Tools 16%
• Non-dominated Solutions 16%
• Vectorized 16%
• Hardware Model 16%
• Non-dominated Sorting Genetic Algorithm (NSGA-II) 16%
• Cost Model 16%
• Three-level 16%
• Globally Optimal 16%
• Hardware Measurement 16%
• Numerical Precision 16%
• Mixed Precision 16%
• Memory-constrained 16%
• Pareto Front Approach 16%
• Neural Network Hardware 16%
• Mixed Precision Quantization 16%
• Discrete Search Space 16%
• Task Interdependence 16%
• Neural Network Combination 16%

Computer Science

• Convolutional Neural Network 100%
• Participatory Design 100%
• Hardware Level 100%
• Computer Hardware 100%
• Search Space 42%
• Target Hardware 28%
• Neural Network Model 14%
• Model Compression 14%
• Multiobjective 14%
• Hardware Design 14%
• Prediction Accuracy 14%
• Embedded Platform 14%
• Varying Degree 14%
• Hardware Architecture 14%
• Hardware Platform 14%
• Genetic Algorithm 14%
• Non-Dominated Sorting Genetic Algorithm II 14%
• Solution Space 14%
• Processing Unit 14%
• Pareto Optimal Set 14%

Engineering

• Abstraction Level 100%
• Convolutional Neural Network 100%
• Search Space 42%
• Design Space 28%
• Processing Unit 14%
• Metrics 14%
• Space Solution 14%
• Pareto Front 14%
• Systematic Approach 14%
• Genetic Algorithm 14%
• Search Method 14%
• Pareto Optimal Set 14%
• Related Task 14%
• Related Cost 14%
• Non-Dominated Sorting Genetic Algorithm II 14%
• Network Model 14%