Provably Accelerating Ill-Conditioned Low-rank Estimation via Scaled Gradient Descent, Even with Overparameterization.
Cong MaXingyu XuTian TongYuejie ChiPublished in: CoRR (2023)
Keyphrases
- low rank
- matrix factorization
- linear combination
- missing data
- convex optimization
- low rank matrix
- singular value decomposition
- rank minimization
- matrix completion
- matrix decomposition
- regularized regression
- high order
- semi supervised
- high dimensional data
- kernel matrix
- cost function
- minimization problems
- trace norm
- singular values
- affinity matrix
- robust principal component analysis
- worst case
- low rank matrices
- low rank approximation
- training data
- feature selection
- non rigid structure from motion
- face recognition
- small number
- higher order
- low dimensional