Toward deeper understanding of neural networks: The power of initialization and a dual view on expressivity

Keyphrases

• Neural Network 100%
• Kernel Space 100%
• Dual-view 100%
• Random Networks 33%
• Activation Function 33%
• Graph Topology 33%
• Nonlinear Transformation 33%
• Random Weights 33%
• Neural Network Training 33%
• Random Neural Network 33%
• Node Replication 33%
• Weight Initialization 33%
• Reproducing Kernel Hilbert Space 33%
• Compositional Transformation 33%
• Acyclic Graph 33%
• Skeleton Graph 33%

Computer Science

• Neural Network 100%
• Deep Understanding 100%
• Starting Point 25%
• Random Network 25%
• Activation Function 25%
• Normal Distribution 25%
• Hilbert Space 25%
• Neural Network Training 25%
• Skeleton Graph 25%
• Nonlinear Transformation 25%

Mathematics

• Neural Network 100%
• Approximates 20%
• Gaussian Distribution 20%
• Worst Case 20%
• Nonlinear 20%
• Starting Point 20%
• Linear Transformation 20%
• Intractability 20%
• Hilbert Space 20%