A regularized Newton method without line search for unconstrained optimization.
Kenji UedaNobuo YamashitaPublished in: Comput. Optim. Appl. (2014)
Keyphrases
- trust region
- newton method
- regularized least squares
- line search
- risk minimization
- quadratic programming
- global convergence
- convergence analysis
- linear equations
- sparse representation
- variational inequalities
- least squares
- convergence rate
- objective function
- loss function
- global optimum
- linear svm
- optimality conditions
- linear programming
- neural network
- convergence speed
- support vector
- reproducing kernel hilbert space
- regularization term
- interior point methods
- cost function
- artificial neural networks