Global Convergence of Newton Method for Empirical Risk Minimization in Reproducing Kernel Hilbert Space.
Ting-Jui ChangShahin ShahrampourPublished in: ACSSC (2020)
Keyphrases
- global convergence
- reproducing kernel hilbert space
- loss function
- global optimum
- convergence rate
- convergence speed
- kernel methods
- optimization methods
- learning theory
- kernel function
- data dependent
- domain adaptation
- special case
- gaussian process
- euclidean space
- density estimation
- distance measure
- real valued
- input space
- learning problems
- metric learning
- search space
- simulated annealing
- support vector machine
- data points
- machine learning
- linear model
- kernel matrix
- pairwise
- support vector