Reexamining the principle of mean-variance preservation for neural network initialization

Keyphrases

• Mean-variance 100%
• Network Initialization 100%
• Neural Network 50%
• Initialize 50%
• Nonlinearity 50%
• Data Dependency 50%
• Sample Variance 50%
• Sample Mean 50%
• Random Networks 50%
• Rectified Linear Unit (ReLU) 50%
• Deep Network 50%
• Backpropagation Training 50%
• Training Intensity 50%

Mathematics

• Neural Network 100%
• Mean-Variance 100%
• Statistics 50%
• Variance 50%
• Random Sample 50%
• Nonlinearities 50%
• Dependent Data 50%
• Sample Variance 50%
• Sample Mean 50%
• Rectification 50%

Economics, Econometrics and Finance

• Nonlinearities 100%