Symmetric block-low-rank layers for fully reversible multilevel neural networks.
Bas PetersEldad HaberKeegan LensinkPublished in: CoRR (2019)
Keyphrases
- low rank
- neural network
- missing data
- matrix factorization
- linear combination
- convex optimization
- low rank matrix
- matrix completion
- rank minimization
- high order
- singular value decomposition
- semi supervised
- high dimensional data
- kernel matrix
- matrix decomposition
- trace norm
- low rank matrices
- robust principal component analysis
- pattern recognition
- low rank representation
- minimization problems
- data matrix
- singular values
- denoising
- affinity matrix
- interior point methods
- wavelet transform