Adaptive Step-Size Methods for Compressed SGD

Keyphrases

• Stochastic Gradient Descent 100%
• Adaptive Step Size 100%
• Stochastic Gradient Algorithm 60%
• Non-adaptive 40%
• Performance Improvement 20%
• Growth Conditions 20%
• Optimal Convergence Rate 20%
• Selection Strategy 20%
• Rule-based 20%
• Convergence Guarantee 20%
• Empirical Performance 20%
• Learning Model 20%
• Distributed Optimization 20%
• Scaling Technique 20%
• Unknown System Parameters 20%
• Theoretical Convergence 20%
• Deep Neural Network 20%
• Order Optimal 20%
• Communication Bottleneck 20%
• Dataset Model 20%
• Federated Learning 20%
• Non-convex Objective 20%
• Interpolation Condition 20%
• Armijo Rule 20%
• Decentralized Optimization Problem 20%
• Step Size Selection 20%

Mathematics

• Stochastics 100%
• Step Size 100%
• Convergence Rate 14%
• System Parameter 14%
• Growth Condition 14%
• Deep Neural Network 14%
• Stochastic Gradient Descent Method 14%
• Interpolation Condition 14%

Computer Science

• Gradient Descent 100%
• Machine Learning 16%
• Experimental Result 16%
• Optimization Problem 16%
• Selection Method 16%
• Convergence Rate 16%
• System Parameter 16%
• Scaling Technique 16%
• Deep Neural Network 16%
• Gradient Descent Method 16%