Optimizing Pipelined Computation and Communication for Latency-Constrained Edge Learning

Mathematics

• Latency 100%
• Communication 69%
• Learning 69%
• Training 40%
• Communication Channels 26%
• Stochastic Gradient 25%
• Gradient Descent 23%
• Machine Learning 20%
• Trade-offs 19%
• Entire 16%
• Optimality 16%
• Regression 14%
• Optimization 14%
• Sufficient 14%
• Numerical Results 12%
• Model 6%

Engineering & Materials Science

• Communication 38%
• Machine learning 27%