Benign Underfitting of Stochastic Gradient Descent

Keyphrases

• Stochastic Gradient Descent 100%
• Underfitting 100%
• Free Replacement 28%
• Population Risk 28%
• Conventional Learning 14%
• Training Data 14%
• Learning Rule 14%
• Optimal Rate 14%
• Generalization Bounds 14%
• Generalization Ability 14%
• Problem Instances 14%
• Generalization Performance 14%
• Optimization Framework 14%
• Multi-epoch 14%
• Uniform Convergence 14%
• Empirical Risk 14%
• Convex Optimization Problem 14%
• Classical Analysis 14%
• Stochastic Convex Optimization 14%
• Finite Sum 14%
• Generalization Gap 14%
• Risk Gap 14%

Computer Science

• Gradient Descent 100%
• Convex Optimization 28%
• Risk Population 28%
• Optimization Problem 14%
• Training Data 14%
• Problem Instance 14%
• Generalization Performance 14%
• Optimization Framework 14%
• Generalization Ability 14%
• Conventional Learning 14%

Mathematics

• Stochastics 100%
• Minimizes 12%
• Main Result 12%
• Lower and upper bounds 12%
• Training Data 12%
• Uniform Convergence 12%
• Learning Rule 12%
• Finite Sum 12%