Exponential Convergence Time of Gradient Descent for One-Dimensional Deep Linear Neural Networks.
Ohad ShamirPublished in: COLT (2019)
Keyphrases
- neural network
- linear complexity
- artificial neural networks
- cost function
- multi dimensional
- operator splitting
- pattern recognition
- update rule
- genetic algorithm
- weight update
- learning rules
- linear constraints
- self organizing maps
- back propagation
- fuzzy logic
- objective function
- convergence rate
- feed forward
- convergence speed
- fuzzy systems
- linear model
- training algorithm
- activation function
- neural network model
- linear systems
- highly non linear