GOALS: Gradient-Only Approximations for Line Searches Towards Robust and Consistent Training of Deep Neural Networks.
Younghwan ChaeDaniel N. WilkeDominic KafkaPublished in: CoRR (2021)
Keyphrases
- neural network
- training process
- training algorithm
- pattern recognition
- feedforward neural networks
- feed forward neural networks
- computationally tractable
- deep architectures
- multi layer perceptron
- artificial neural networks
- neural network training
- backpropagation algorithm
- genetic algorithm
- fuzzy logic
- edge detection
- online learning
- error back propagation
- recurrent networks
- data sets
- photometric invariant
- radial basis function network
- feed forward
- neural network model
- line segments
- training samples
- computationally efficient
- least squares
- search engine