Toward the Optimal Construction of a Loss Function Without Spurious Local Minima for Solving Quadratic Equations.
Zhenzhen LiJian-Feng CaiKe WeiPublished in: IEEE Trans. Inf. Theory (2020)
Keyphrases
- loss function
- pairwise
- support vector
- empirical risk
- nonlinear equations
- learning to rank
- solution path
- logistic regression
- boosting framework
- risk minimization
- boosting algorithms
- hinge loss
- optimal solution
- regularization term
- global minimum
- reproducing kernel hilbert space
- simulated annealing
- worst case
- special case
- lower bound