Generalization Performance of Multi-pass Stochastic Gradient Descent with Convex Loss Functions

Keyphrases

• Generalization Performance 100%
• Stochastic Gradient Descent 100%
• Convex Loss Function 100%
• Multi-pass 100%
• Learning Rate 80%
• Smoothness Assumption 40%
• Subgradient 40%
• Existing Capacity 40%
• Implicit Regularization 20%
• Practical Performance 20%
• Empirical Process 20%
• Computational Properties 20%
• Low Computational Complexity 20%
• High Probability 20%
• Rate Analysis 20%
• Large-scale Dataset 20%
• Statistical Properties 20%
• Martingale 20%
• Choice Method 20%
• Independent Learning 20%
• Concentration Inequalities 20%
• Least Squares Loss 20%
• Number of Passes 20%
• Optimal Capacity 20%

Mathematics

• Stochastics 100%
• Loss Function 100%
• Subgradient 40%
• Least Square 20%
• Probability Theory 20%
• Regularization 20%
• Computational Cost 20%
• Empirical Process 20%
• Statistical Property 20%
• Square Loss 20%