AdaLoss: A computationally-efficient and provably convergent adaptive gradient method.
Xiaoxia WuYuege XieSimon S. DuRachel WardPublished in: CoRR (2021)
Keyphrases
- gradient method
- computationally efficient
- provably convergent
- convergence rate
- shape from shading
- recursive least squares
- negative matrix factorization
- image processing
- convex formulation
- step size
- optimization methods
- image classification
- image segmentation
- image representation
- image compression
- convergence speed
- support vector machine
- computational complexity
- clustering algorithm
- computer vision
- natural gradient learning