Implicit Regularization of Discrete Gradient Dynamics in Deep Linear Neural Networks.
Gauthier GidelFrancis R. BachSimon Lacoste-JulienPublished in: CoRR (2019)
Keyphrases
- neural network
- global exponential stability
- pattern recognition
- back propagation
- dynamical systems
- steepest descent method
- fuzzy logic
- nonlinear dynamics
- hopfield neural network
- linear systems
- continuous variables
- neural nets
- dynamic model
- feed forward
- neural network model
- closed form
- linear functions
- regularization parameter
- piecewise linear
- edge detection
- artificial neural networks
- highly nonlinear
- binary valued
- gauss seidel
- multi layer
- deep learning
- training process
- fuzzy systems
- self organizing maps
- semi supervised
- genetic algorithm