Random Reshuffling with Variance Reduction: New Analysis and Better Rates

Mathematics

• Variance Reduction 100%
• Stochastics 100%
• Training Data 75%
• Convergence Rate 75%
• Linear Convergence 50%
• Sampling Without Replacement 50%
• Data Point 50%
• Stochastic Gradient Descent Method 25%
• Number 25%

Computer Science

• Gradient Descent 100%
• Variance Reduction 100%
• Convergence Rate 75%
• Machine Learning 50%
• Training Data Point 50%
• Training Data 25%
• Gradient Descent Method 25%
• Reduction Mechanism 25%

Keyphrases

• Sampling without Replacement 50%
• Random Shuffling 25%
• Data Order 25%