Linearization Weight Compression and In-Situ Hardware-Based Decompression for Attention-Based Neural Machine Translation

Engineering

• Decompression 100%
• Compression Method 100%
• Energy Engineering 50%
• Quantization (Signal Processing) 50%
• Energy Conservation 50%
• Fixed Points 50%
• Acceleration Method 50%
• Data Transfer 50%
• Multi-Head Attention 50%
• Compression Ratio 50%

Computer Science

• Machine Translation 100%
• translation model 44%
• Bidirectional Encoder Representations From Transformers 33%
• Latency Reduction 11%
• Quantization (Signal Processing) 11%
• Energy Consumption 11%
• Energy Saving 11%
• hardware acceleration 11%
• Compression Ratio 11%
• Fixed Points 11%
• Data Memory 11%