On the Proof of Global Convergence of Gradient Descent for Deep ReLU Networks with Linear Widths.
Quynh NguyenPublished in: ICML (2021)
Keyphrases
- global convergence
- global optimum
- convergence rate
- convergence speed
- convergence analysis
- optimization methods
- coordinate ascent
- objective function
- constrained optimization problems
- convex minimization
- line search
- loss function
- social networks
- network structure
- cost function
- conjugate gradient
- globally convergent
- multiscale
- search algorithm
- machine learning
- linear svm
- multi objective
- linear constraints
- step size
- least squares
- linear combination
- hybrid algorithm