Low-Rank Gradient Approximation for Memory-Efficient on-Device Training of Deep Neural Network.
Mary GooneratneKhe Chai SimPetr ZadrazilAndreas KabelFrançoise BeaufaysGiovanni MottaPublished in: ICASSP (2020)
Keyphrases
- memory efficient
- low rank
- neural network
- matrix factorization
- training process
- matrix decomposition
- missing data
- low rank matrix
- convex optimization
- linear combination
- matrix completion
- singular value decomposition
- semi supervised
- low rank approximation
- rank minimization
- high dimensional data
- kernel matrix
- approximation methods
- non rigid structure from motion
- robust principal component analysis
- training samples
- high order
- low rank matrices
- minimization problems
- pattern recognition
- collaborative filtering
- singular values
- data matrix
- stochastic gradient descent
- trace norm
- data mining
- image processing
- training data
- multiscale
- training set
- small number
- least squares
- supervised learning