Optimization and generalization of shallow neural networks with quadratic activation functions

Keyphrases

• Neural Network 100%
• Activation Function 100%
• Gradient Descent 100%
• Hidden Layer 100%
• Shallow Neural Network 100%
• Student Teachers 50%
• Sampling numbers 50%
• Convergence Rate 50%
• Numerical Experiments 50%
• Generalization Properties 50%
• Teacher Networks 50%
• Generalization Error 50%
• Loss Landscape 50%
• Layer Width 50%

Mathematics

• Neural Network 100%
• Probability Theory 33%
• Numerical Experiment 33%
• Data Sample 33%
• Convergence Rate 33%

Chemical Engineering

• Neural Network 100%