Energy-Efficient Recurrent Neural Network With MRAM-Based Probabilistic Activation Functions

Keyphrases

• Energy Efficient 100%
• Recurrent Neural Network 100%
• Activation Function 100%
• Hardware Complexity 33%
• Energy Efficiency 33%
• Synapse 33%
• Low Power 33%
• Proposed Design 33%
• Matching Algorithm 33%
• Resistive Random Access Memory (ReRAM) 33%
• Ultra-low Power 33%
• Efficient Hardware Implementation 33%
• Order of Accuracy 33%
• Activation Mechanism 33%
• Data Parallelism 33%
• Tanh 33%
• High Energy Efficiency 33%
• Area-efficient 33%
• Energy-efficient Hardware 33%
• Cross-layer Model 33%
• Processing-In-Memory Architectures 33%
• Evaluation Phase 33%
• Area Reduction 33%

Computer Science

• Energy Efficient 100%
• Recurrent Neural Network 100%
• Activation Function 100%
• Energy Efficiency 66%
• Random Access Memory 33%
• Matching Algorithm 33%
• Level Parallelism 33%
• Hardware Implementation 33%
• Memory Architecture 33%
• Evaluation Phase 33%