Fast Convergence of Natural Gradient Descent for Overparameterized Neural Networks.
Guodong ZhangJames MartensRoger B. GrossePublished in: CoRR (2019)
Keyphrases
- neural network
- weight update
- update rule
- learning rules
- convergence rate
- pattern recognition
- genetic algorithm
- evolutionary algorithm
- multilayer perceptron
- back propagation
- feed forward
- conjugate gradient
- real world
- cost function
- artificial neural networks
- fault diagnosis
- image restoration
- convergence speed
- network architecture
- initial conditions
- iterative algorithms
- artificial systems
- adaptive resonance theory
- objective function
- real time