Learning One-hidden-layer ReLU Networks via Gradient Descent

Keyphrases

• Hidden Layer 100%
• Gradient Descent 100%
• Rectified Linear Unit (ReLU) 100%
• Neural Network 66%
• Statistical Errors 33%
• Activation Function 33%
• Numerical Experiments 33%
• Teacher Networks 33%
• Gaussian Distribution 33%
• Linear Growth Rate 33%
• Empirical Risk Minimization 33%
• Recovery Guarantee 33%
• Practical Learning 33%

Mathematics

• Neural Network 100%
• Gaussian Distribution 50%
• Tensor 50%
• Statistical Error 50%
• Numerical Experiment 50%
• Empirical Risk Minimization 50%

Computer Science

• Gradient Descent 100%
• ReLU Function 100%
• Neural Network 33%
• Activation Function 33%
• Risk Minimization 33%
• Layer Neural Network 33%