PowerSGD: Practical Low-Rank Gradient Compression for Distributed Optimization.
Thijs VogelsSai Praneeth KarimireddyMartin JaggiPublished in: CoRR (2019)
Keyphrases
- low rank
- nuclear norm
- convex optimization
- missing data
- linear combination
- matrix factorization
- rank minimization
- matrix completion
- low rank matrix
- singular value decomposition
- matrix decomposition
- kernel matrix
- high order
- semi supervised
- image compression
- low rank matrices
- minimization problems
- convex relaxation
- singular values
- trace norm
- high dimensional data
- data mining
- robust principal component analysis
- low rank and sparse
- input data
- pattern recognition
- image processing