Towards Execution-Efficient LSTMs via Hardware-Guided Grow-and-Prune Paradigm

Computer Science

• Efficient Execution 100%
• Long Short-Term Memory Network 100%
• Execution Efficiency 50%
• Inference Model 50%
• Neural Network 25%
• Sparsity 25%
• Memory Architecture 25%
• Speech Recognition 25%
• Memory Latency 25%
• Language Modeling 25%
• Complexity Reduction 25%
• Model Architecture 25%
• Word Error Rate 25%

Keyphrases

• Long Short-term Memory Architecture 50%
• Training Methodology 50%
• Stacked Long Short-term Memory 50%