HAS-RL: A Hierarchical Approximate Scheme Optimized With Reinforcement Learning for NoC-Based NN Accelerators

Computer Science

• Networks on Chips 100%
• Reinforcement Learning 100%
• Communication Delay 33%
• Application 33%
• Approximation (Algorithm) 33%
• Neural Network 33%
• Accuracy 33%
• Machine Learning 16%
• Dynamic Traffic 16%
• State Space 16%
• Design 16%
• Trained Network 16%
• Search Space 16%
• Power Consumption 16%
• Complexity 16%
• quality model 16%
• Routers 16%
• Transmitted Data 16%
• Communication 16%
• Communication Architecture 16%
• Relationships 16%
• Traffic Pattern 16%

INIS

• approximations 100%
• accelerators 100%
• learning 100%
• space 23%
• data 23%
• applications 15%
• losses 15%
• neural networks 15%
• comparative evaluations 15%
• accuracy 15%
• solutions 15%
• design 7%
• dynamics 7%
• power 7%
• regulations 7%
• resources 7%
• machine learning 7%
• policy 7%
• architecture 7%
• performance 7%
• algorithms 7%
• balances 7%
• constraints 7%

Keyphrases

• Average Network Delay 20%
• On-chip Communication Architecture 20%
• Router Design 20%
• Approximate Communication 20%