Minnorm training: an algorithm for training over-parameterized deep neural networks.
Yamini BansalMadhu AdvaniDavid D. CoxAndrew M. SaxePublished in: CoRR (2018)
Keyphrases
- training process
- training algorithm
- neural network
- neural network training
- learning algorithm
- neural network structure
- computational complexity
- times faster
- dynamic programming
- feed forward neural networks
- training samples
- optimization algorithm
- convergence rate
- training phase
- learning speed
- training procedure
- training speed
- early stopping
- objective function
- preprocessing
- computational cost
- particle swarm optimization
- search space
- k means
- probabilistic model
- pattern recognition
- radial basis function network
- feedforward neural networks
- segmentation algorithm
- fuzzy logic
- supervised learning
- worst case
- simulated annealing
- training examples
- detection algorithm
- expectation maximization