Computational Complexity of Learning Neural Networks: Smoothness and Degeneracy

Keyphrases

• Computational Complexity 100%
• Deep Network 25%
• Gaussian Distribution 25%
• Gaussian Input 25%
• Hardness of Learning 25%
• Hardness Results 50%
• Input Distribution 100%
• Learning Complexity 100%
• Learning Neural Network 100%
• Learning Theory 25%
• Network Parameters 50%
• Network Weight 25%
• Neural Network 25%
• Non-degenerate 50%
• Pseudorandom number Generator 25%
• Random Noise 25%
• ReLU Networks 50%
• Smoothed Analysis 50%
• Weight Matrix 50%

Computer Science

• Analysis Framework 50%
• Computational Complexity 100%
• Deep Learning Method 25%
• Efficient Algorithm 25%
• Fundamental Question 25%
• Gaussian Input 25%
• Input Distribution 100%
• Network Parameter 50%
• Neural Network 100%

Mathematics

• Deep Learning Method 33%
• Gaussian Distribution 100%
• Network Weight 33%
• Neural Network 100%
• Random Noise 33%
• Weight Matrix 66%