Eigencurve: Optimal Learning Rate Schedule for SGD on Quadratic Objectives with Skewed Hessian Spectrums.
Rui PanHaishan YeTong ZhangPublished in: CoRR (2021)
Keyphrases
- learning rate
- convergence rate
- rapid convergence
- training speed
- scheduling problem
- error function
- dynamic programming
- multilayer neural networks
- optimal solution
- objective function
- hidden layer
- convergence speed
- learning algorithm
- adaptive learning rate
- convergence theorem
- bp neural network algorithm
- step size
- weight vector
- high accuracy
- worst case
- feature selection