Gradient regularization of Newton method with Bregman distances.
Nikita DoikovYurii E. NesterovPublished in: Math. Program. (2024)
Keyphrases
- newton method
- regularized least squares
- regularization term
- total variation
- convergence analysis
- bregman divergences
- linear equations
- optimality conditions
- nonnegative matrix factorization
- variational inequalities
- loss function
- reproducing kernel hilbert space
- distance measure
- linear svm
- quadratic programming
- regularization parameter
- data dependent
- global convergence
- convergence rate
- image processing
- logistic regression
- support vector