AUTOQ: AUTOMATED KERNEL-WISE NEURAL NETWORK QUANTIZATION

Keyphrases

• Inference Accuracy 100%
• Neural Network Quantization 100%
• Convolutional Neural Network 75%
• Learning-based 50%
• Redundancy 50%
• Quantization Technique 50%
• Deep Reinforcement Learning (deep RL) 50%
• Convolutional Layer 50%
• Low Power 25%
• Design Space 25%
• Hardware Efficiency 25%
• Mobile Devices 25%
• Gradient-based 25%
• Energy Consumption 25%
• Energy Overhead 25%
• Hardware Overhead 25%
• Suboptimal Outcome 25%
• Latency Overhead 25%
• Deep Deterministic Policy Gradient 25%
• Neural Network Inference 25%
• Inference Latency 25%
• Activation Layer 25%
• Hierarchical Deep Reinforcement Learning 25%

Computer Science

• Neural Network 100%
• Deep Reinforcement Learning 100%
• Convolutional Neural Network 100%
• Convolutional Layer 66%
• Energy Consumption 33%
• Hardware Overhead 33%
• Network Inference 33%
• Computer Hardware 33%
• Mobile Device 33%
• Activation Layer 33%