On the Global Convergence of Gradient Descent for multi-layer ResNets in the mean-field regime.
Zhiyan DingShi ChenQin LiStephen J. WrightPublished in: CoRR (2021)
Keyphrases
- multi layer
- global convergence
- coordinate ascent
- global optimum
- convergence rate
- convergence speed
- optimization methods
- convergence analysis
- objective function
- cost function
- neural network
- conjugate gradient
- markov random field
- neural nets
- loss function
- step size
- convex minimization
- em algorithm
- gauss newton
- particle swarm
- multiple layers
- hybrid algorithm
- differential evolution
- natural images
- particle swarm optimization
- search space
- multiscale