Neural Network Training with Safe Regularization in the Null Space of Batch Activations.
Matthias KisselMartin GottwaldKlaus DiepoldPublished in: ICANN (2) (2020)
Keyphrases
- neural network training
- null space
- neural network
- training algorithm
- linear discriminant analysis
- novelty detection
- singular value decomposition
- principal components
- optimization method
- singular values
- discriminative information
- discriminant analysis
- lower dimensional
- face recognition
- theoretical guarantees
- finite dimensional
- particle swarm optimisation
- dimensionality reduction
- intra class
- principal component analysis
- latent variables
- small sample size
- support vector
- k nearest neighbor