error of an over-parametrized deep neural network estimate learned by gradient descent without regularization.
Selina DrewsMichael KohlerPublished in: CoRR (2023)
Keyphrases
- neural network
- estimation error
- error function
- cost function
- pattern recognition
- loss function
- self organizing maps
- recurrent neural networks
- neural network is trained
- training algorithm
- fuzzy logic
- artificial neural networks
- back propagation
- learning rules
- objective function
- stochastic gradient descent
- estimated parameters
- fault diagnosis
- prior information
- multi layer
- regularization term
- activation function
- deep learning
- image processing